Advancing Precision Medicine: Developing Polygenic Risk Scores for Endometriosis Subphenotypes

Abstract

Endometriosis is a complex gynecological disorder affecting 6-10% of reproductive-aged women, characterized by significant diagnostic delays of 7-12 years. This article explores the development and application of polygenic risk scores (PRS) for endometriosis subphenotypes to enable earlier detection and personalized treatment approaches. We review foundational genetic discoveries from genome-wide association studies that have identified multiple risk loci, particularly for moderate-to-severe disease. The content covers methodological advances in PRS construction, current challenges in clinical prediction, and emerging strategies that integrate epigenetic data such as methylation risk scores. For researchers and drug development professionals, we provide a comprehensive analysis of validation frameworks and comparative performance against traditional risk factors, highlighting future directions for implementing genetic risk stratification in clinical practice and therapeutic development.

The Genetic Architecture of Endometriosis: From GWAS to Subphenotype Differentiation

Endometriosis is a complex, chronic inflammatory gynecological condition characterized by the presence of endometrial-like tissue outside the uterine cavity, affecting approximately 10% of women of reproductive age worldwide [1]. Its etiology involves a multifactorial interplay of genetic, hormonal, immune, and environmental factors. Establishing its heritability and genetic architecture is a critical foundation for developing polygenic risk scores (PRS) capable of stratifying disease risk and subphenotypes, ultimately advancing personalized medicine approaches for this heterogeneous condition [1] [2].

Family and twin studies provide the fundamental evidence for a significant genetic component in endometriosis. Family studies demonstrate a five- to seven-fold increased risk for first-degree relatives of affected individuals compared to the general population [2]. Twin studies reveal higher concordance rates in monozygotic twins compared to dizygotic twins, with estimated heritability reaching up to 50% based on genome-wide association studies (GWAS) and linkage analyses [1] [2]. Furthermore, familial cases often present with an earlier onset and more severe symptoms than sporadic cases, suggesting a potentially greater genetic burden in these families [2].

Table 1: Key Evidence of Heritability in Endometriosis

Evidence Type	Key Finding	Implication for Genetics
Family Studies	5-7x increased risk for first-degree relatives [2]	Strong evidence for inherited genetic components
Twin Studies	Higher concordance in identical twins; heritability ~50% [1] [2]	Indicates significant genetic contribution, separate from shared environment
Familial Case Presentation	Earlier onset and more severe symptoms [2]	Suggests a higher genetic burden or different genetic architecture

Established Genetic Risk Architecture

Common Variants from Genome-Wide Association Studies (GWAS)

GWAS have successfully identified multiple common, low-penetrance genetic variants associated with endometriosis risk. These studies have identified single nucleotide polymorphisms (SNPs) in genes often involved in sex steroid hormone pathways, including WNT4, VEZT, GREB1, ESR1, and FSHB [1] [2]. These common variants individually confer modest risk increases, but in combination, they account for a portion of the disease's heritability, supporting the polygenic nature of endometriosis.

Rare Variants and Familial Clustering

Despite the success of GWAS, a substantial fraction of heritability remains unexplained, prompting investigations into the role of rare, higher-penetrance variants, particularly in multi-affected families. A recent exploratory whole-exome sequencing (WES) study of a multigenerational family with multiple affected members identified 36 co-segregating rare variants [2]. The top candidate genes from this study were LAMB4 (c.3319G>A, p.Gly1107Arg) and EGFL6 (c.1414G>A, p.Gly472Arg), which are associated with cancer growth and tissue remodeling. Variants in NAV3, ADAMTS18, SLIT1, and MLH1 were also identified as potential contributors, supporting a polygenic or oligogenic model where multiple rare variants act synergistically to increase disease susceptibility in familial cases [2].

Table 2: Summary of Key Genetic Findings in Endometriosis

Genetic Element	Examples	Method of Discovery	Biological Implication
Common Variants (SNPs)	WNT4, VEZT, GREB1, ESR1, FSHB [1] [2]	GWAS	Hormone signaling, cellular growth and maintenance
Rare Variants (Candidate)	LAMB4, EGFL6 [2]	Whole-Exome Sequencing (Familial)	Cell adhesion, extracellular matrix formation, angiogenesis
Epigenetic Alterations	DNA methylation of estrogen metabolism genes; miRNA dysregulation [1] [2]	Epigenomic Studies	Altered gene expression contributing to estrogen dominance and progesterone resistance

Methodologies for Establishing Genetic Burden

Family-Based Whole-Exome Sequencing

Objective: To identify rare, penetrant genetic variants that co-segregate with endometriosis in multi-affected families.

Workflow:

• Family Selection: Identify and recruit a multigenerational family with multiple affected individuals (e.g., three sisters, their mother, and grandmother) [2].
• Sample Collection: Obtain peripheral blood from affected family members.
• DNA Extraction & WES: Extract genomic DNA from leukocytes and perform whole-exome sequencing using a platform like Illumina with an average coverage of >100x [2].
• Bioinformatic Analysis:
  • Read Mapping: Map sequence reads to a reference genome (e.g., GRCh37/hg19) using BWA.
  • Variant Calling: Identify variants using a caller like FreeBayes. Each individual typically yields ~20,000-25,000 raw variants.
  • Quality Control (QC): Apply filters for depth, genotype quality, and call rate, reducing variants to ~15,000-20,000 per individual.
  • Variant Filtering: Focus on rare (e.g., MAF < 0.01), protein-altering variants (missense, frameshift, stop-gain) that co-segregate with the disease in affected members.
• Prioritization & Validation: Prioritize candidate genes based on predicted pathogenicity and known biological functions. Replication in independent cohorts and functional studies are required for validation [2].

Genome-Wide Association Study (GWAS) for PRS Development

Objective: To identify common genetic variants associated with endometriosis risk and generate summary statistics for polygenic risk score calculation.

Workflow:

• Cohort Ascertainment: Assemble large, independent case-control cohorts with stringent phenotyping (e.g., surgical confirmation).
• Genotyping & Imputation: Genotype all participants using a microarray. Impute to a reference panel (e.g., 1000 Genomes) to increase the number of testable SNPs.
• Association Analysis: Perform a logistic regression for each SNP, adjusting for principal components to control for population stratification.
• Meta-Analysis: Combine summary statistics from multiple GWAS to increase power.
• PRS Calculation: The resulting GWAS summary statistics serve as the "base data" for PRS calculation. The score for an individual in a target dataset is computed as the weighted sum of their risk alleles: PRS = β₁SNP₁ + β₂SNP₂ + ... + βₙSNPₙ, where β is the effect size from the GWAS [3].
• PRS Validation: The PRS must be validated for association and predictive performance in independent target datasets [3].

Quality Control for PRS Analysis

Robust PRS analysis requires stringent QC on both base (GWAS) and target datasets [3]:

• Base Data QC: Ensure SNP heritability (h²SNP) > 0.05; verify effect allele identity; use consistent genome build.
• Target Data QC: Use a target sample size ≥100; perform standard GWAS QC; remove ambiguous and duplicate SNPs; exclude sex chromosomes if not relevant; remove samples overlapping with the base GWAS and closely related individuals.
• Data Integration: Strand-flip alleles to resolve mismatches; use clumping or LD-adjustment methods to account for linkage disequilibrium.

The Scientist's Toolkit: Research Reagent Solutions

Table 3: Essential Research Materials and Reagents

Item / Reagent	Function / Application	Example Use in Context
Illumina WGS/WES Platform	High-throughput sequencing to identify genetic variants.	Germline variant discovery in multi-affected families [2].
SOMAscan Proteomic Platform	Multiplexed immunoaffinity assay to measure plasma protein levels (pQTLs).	Identifying protein biomarkers and therapeutic targets via Mendelian randomization [4].
Human R-Spondin3 (RSPO3) ELISA Kit	Quantitatively measure protein concentration in plasma.	Validating RSPO3 as a potential therapeutic target in patient plasma samples [4].
Galaxy Platform	Web-based platform for accessible, reproducible bioinformatic analysis.	Processing WES data (read mapping, duplicate removal, variant calling) [2].
PLINK Software	Whole-genome association analysis toolset.	Performing LD clumping and basic QC for PRS calculation [3].

Advanced Analytical Frameworks

Mendelian Randomization for Target Discovery

Objective: To assess causal relationships between putative risk factors (e.g., plasma proteins, metabolites) and endometriosis using genetic variants as instrumental variables.

Workflow:

• Instrument Selection: Select genetic variants (e.g., cis-pQTLs) strongly associated (P < 5×10⁻⁸) with the exposure (e.g., RSPO3 protein levels), ensuring they are independent of confounders and influence the outcome only via the exposure [4].
• Data Sources: Obtain summary statistics from large-scale GWAS for the exposure and outcome (endometriosis).
• MR Analysis: Apply methods (e.g., Inverse-Variance Weighted) to estimate the causal effect.
• Validation: Use colocalization analysis (e.g., calculating posterior probability of shared causal variant, PPH4) to confirm robust associations. A recent study identified RSPO3 as a potential causal protein and therapeutic target for endometriosis using this approach [4].

Artificial Intelligence in Genomic Prediction

Machine learning and deep learning models are increasingly applied to enhance genomic prediction of complex diseases like endometriosis. These models can capture non-linear effects and complex interactions between genetic variants that are missed by traditional linear PRS methods [5] [3]. For instance, a multi-variant deep neural network (DNN) approach has been explored to improve the genomic prediction of endometriosis, demonstrating the potential of AI to handle the high-dimensional nature of genomic data and integrate it with other clinical risk factors for more accurate risk stratification [5].

Evidence from twin and family studies unequivocally establishes a substantial genetic component in endometriosis, with a heritability estimate of approximately 50%. Its genetic architecture is complex, involving a spectrum of variants from common, low-penetrance SNPs identified by GWAS to rare, potentially higher-penetrance variants discovered in familial cases. Methodologies like family-based WES, large-scale GWAS, and advanced analytical frameworks such as Mendelian randomization and AI-driven modeling are critical for dissecting this burden. A comprehensive understanding of this genetic landscape is the essential foundation for developing next-generation polygenic risk scores that can stratify subphenotypes and drive forward personalized therapeutic strategies and preventive care for endometriosis.

Endometriosis is a common, estrogen-dependent inflammatory gynecological disorder that affects approximately 10% of women of reproductive age, representing over 190 million women worldwide [6] [7]. The disease is characterized by the presence of endometrial-like tissue outside the uterine cavity and is associated with chronic pelvic pain, reduced fertility, and decreased quality of life [8]. The heritability of endometriosis is estimated to be 47-52%, indicating a strong genetic component [8] [9]. Genome-wide association studies (GWAS) have emerged as a powerful hypothesis-free approach for identifying common genetic variants underlying complex diseases like endometriosis. This application note summarizes key GWAS discoveries of genetic loci associated with overall endometriosis risk, framed within the context of developing polygenic risk scores for endometriosis subphenotypes.

Key GWAS Discoveries and Associated Loci

Over the past decade, multiple GWAS and meta-analyses have substantially expanded our understanding of the genetic architecture of endometriosis. The largest initial GWAS meta-analysis published in 2017 analyzed 17,045 cases and 191,596 controls of European and Japanese ancestry, identifying 19 independent single nucleotide polymorphisms (SNPs) robustly associated with endometriosis risk [10]. These SNPs together explained approximately 5.19% of the disease variance, highlighting the highly polygenic nature of endometriosis [10]. More recent combinatorial analytics approaches have identified additional multi-SNP disease signatures, comprising 2,957 unique SNPs in combinations of 2-5 SNPs, that were associated with increased prevalence of endometriosis [6].

Table 1: Key Endometriosis Risk Loci Identified through GWAS

Genomic Region	Lead SNP	Nearest Gene(s)	Reported OR	P-value	Primary Biological Pathway
1p36.12	rs7521902	WNT4	1.11	1.8 × 10-15	Reproductive development, hormone signaling
2p25.1	rs13391619	GREB1	1.09	4.5 × 10-8	Estrogen regulation, cell proliferation
6q25.1	rs71575922	SYNE1, ESR1	1.11	2.02 × 10-8	Sex steroid hormone signaling
7p15.2	rs12700667	Intergenic	1.12	1.6 × 10-9	Inflammatory response
9p21.3	rs1537377	CDKN2B-AS1	1.14	1.5 × 10-8	Cell cycle regulation
12q22	rs10859871	VEZT	1.12	4.7 × 10-15	Cell adhesion
11p14.1	rs74485684	FSHB	1.11	2.00 × 10-8	Gonadotropin hormone production
2q35	rs1250241	FN1	1.23	2.99 × 10-9	Tissue remodeling, fibrosis
6q25.1	rs1971256	CCDC170	1.09	3.74 × 10-8	Estrogen receptor signaling

Notably, most endometriosis risk loci discovered through GWAS are located in non-coding regions of the genome, suggesting they likely influence gene regulation rather than protein structure [8]. Integration of GWAS findings with expression quantitative trait loci (eQTL) data from physiologically relevant tissues (uterus, ovary, vagina, colon, ileum, and peripheral blood) has provided insights into the functional consequences of these variants [7].

Experimental Protocols for Endometriosis GWAS

Standard GWAS Protocol

Objective: To identify genetic variants associated with endometriosis risk through a genome-wide case-control association study.

Materials:

• DNA samples from endometriosis cases and controls
• High-density SNP genotyping arrays
• Quality control tools (PLINK, EIGENSOFT)
• Imputation software (IMPUTE2, Minimac)
• Association analysis software (SNPTEST, PLINK)

Procedure:

• Sample Collection and Diagnosis: Recruit surgically confirmed endometriosis cases (preferably with rAFS staging) and age-matched controls without endometriosis.
• Genotype Data Generation: Extract genomic DNA and genotype using genome-wide SNP arrays (e.g., Illumina Global Screening Array).
• Quality Control:
  • Sample QC: Exclude samples with call rate <95%, sex mismatches, excessive heterozygosity, or divergent ancestry.
  • Variant QC: Exclude SNPs with call rate <95%, Hardy-Weinberg equilibrium p<1×10^-6 in controls, or minor allele frequency <1%.
• Population Stratification: Apply principal component analysis to identify and control for population structure.
• Genotype Imputation: Perform phasing and imputation using reference panels (1000 Genomes Project or HRC) to increase SNP density.
• Association Testing: Conduct logistic regression assuming an additive genetic model, adjusting for principal components.
• Meta-analysis: Combine results across multiple studies using fixed-effects models (e.g., METAL software).
• Significance Threshold: Apply genome-wide significance threshold of p<5×10^-8.

Combinatorial Analytics Protocol

Objective: To identify multi-SNP combinations associated with endometriosis risk using combinatorial analytics.

Materials:

• Genotype data from UK Biobank or other biobanks
• PrecisionLife combinatorial analytics platform or similar tool
• Statistical computing environment (R, Python)

Procedure:

• Data Preparation: Curate endometriosis cases and controls from biobank resources, ensuring appropriate phenotyping.
• Combinatorial Analysis: Use the combinatorial analytics platform to analyze SNP combinations of 2-5 variants for association with endometriosis prevalence.
• Validation: Test significant multi-SNP signatures in independent validation cohorts (e.g., All of Us cohort).
• Pathway Enrichment: Perform pathway analysis on genes mapped to reproducing signatures using databases like MSigDB.
• Novel Gene Identification: Prioritize novel genes occurring in high-frequency reproducing signatures without linkage to known GWAS hits.

Signaling Pathways and Biological Mechanisms

Integration of GWAS findings with functional genomic data has elucidated key biological pathways involved in endometriosis pathogenesis. The diagram below illustrates the major signaling pathways through which GWAS-identified genetic loci contribute to endometriosis risk.

Diagram 1: Signaling pathways connecting GWAS-identified loci to endometriosis pathogenesis. Genetic variants influence disease risk through hormonal signaling, immune regulation, and cellular processes.

Research Reagent Solutions

Table 2: Essential Research Reagents for Endometriosis Genetic Studies

Reagent/Material	Function/Application	Example Specifications
High-Density SNP Arrays	Genome-wide genotyping	Illumina Global Screening Array (700,000+ markers)
Whole Genome Sequencing Kits	Comprehensive variant detection	Illumina NovaSeq, PacBio HiFi for structural variants
DNA Extraction Kits	High-quality DNA isolation from blood/tissue	QIAamp DNA Blood Maxi Kit (Qiagen)
eQTL Reference Datasets	Functional annotation of risk variants	GTEx v8 ( uterus, ovary, blood tissues)
Pathway Analysis Software	Biological interpretation of GWAS hits	GSEA-MSigDB, Ingenuity Pathway Analysis
Genotype Imputation Services	Increased SNP coverage from array data	Michigan Imputation Server (TOPMed reference)
Cell Line Models	Functional validation of risk genes	Endometrial stromal cells, epithelial organoids
CRISPR-Cas9 Systems	Gene editing for functional studies	Lentiviral CRISPR libraries for high-throughput screening

Implications for Polygenic Risk Score Development

The GWAS discoveries summarized herein provide the foundation for developing polygenic risk scores (PRS) for endometriosis. A PRS derived from 14 genome-wide significant variants has demonstrated association with endometriosis in multiple cohorts, with odds ratios ranging from 1.28 to 1.59 per standard deviation increase in PRS [11]. Importantly, the PRS was associated with all major subtypes of endometriosis (ovarian, infiltrating, and peritoneal) but not with adenomyosis, suggesting specificity for endometriosis rather than general gynecological pathology [11].

Recent PRS-phenome-wide association studies have revealed pleiotropic effects of endometriosis genetic risk, including associations with lower testosterone levels, suggesting potential causal relationships [9]. Combinatorial analytics approaches have identified additional multi-SNP signatures that show high reproducibility (73-85%) across diverse ancestries, providing enhanced resolution for subtype-specific genetic architecture [6].

The continuing expansion of GWAS discoveries, including recent multi-ancestry studies encompassing ~1.4 million women, will further refine PRS development and enable more precise stratification of endometriosis subphenotypes [12]. Integration of functional genomic data with GWAS findings will facilitate the translation of genetic discoveries into pathogenic mechanisms and therapeutic targets, ultimately enabling precision medicine approaches for this complex disorder.

Application Note: Genetic Landscape of Endometriosis Subphenotypes

Endometriosis is a heterogeneous gynecological condition affecting approximately 10% of reproductive-aged women globally, characterized by the presence of endometrial-like tissue outside the uterine cavity [1] [13]. The disease manifests in distinct subphenotypes including ovarian endometriosis (endometriomas), deep infiltrating endometriosis (DIE), and superficial peritoneal endometriosis (SPE), each demonstrating unique clinical presentations and molecular characteristics [14]. Understanding the genetic architecture underlying these subphenotypes is crucial for developing polygenic risk scores (PRS) with improved predictive accuracy and clinical utility. This application note synthesizes current evidence on subphenotype-specific genetic associations and provides methodological frameworks for PRS development in endometriosis research.

Subphenotype Classification and Clinical Characteristics

Endometriosis subphenotypes are classified based on lesion location, invasiveness, and histological features [14]. Ovarian endometriosis presents as cystic lesions (endometriomas) containing dark, chocolate-colored fluid. Deep infiltrating endometriosis penetrates more than 5 mm beneath the peritoneal surface and can involve uterosacral ligaments, rectovaginal septum, bowel, bladder, and ureters. Superficial peritoneal endometriosis appears as superficial implants on pelvic peritoneum. A recent classification system stages genital endometriosis from minimal (Stage I) to severe (Stage IV) based on lesion number, penetration depth, adhesion presence, and concomitant adenomyosis [14].

Table 1: Clinical and Pathological Features of Endometriosis Subphenotypes

Subphenotype	Lesion Characteristics	Common Locations	Invasiveness	Associated Symptoms
Ovarian Endometriosis	Cystic lesions (endometriomas) containing old blood	Ovaries	Non-infiltrating, expansive growth	Pelvic pain, dysmenorrhea, infertility
Deep Infiltrating Endometriosis (DIE)	Solid lesions with >5mm penetration depth	Rectovaginal septum, uterosacral ligaments, bowel, bladder, ureters	Highly infiltrative	Severe chronic pelvic pain, dyspareunia, dyschezia, infertility
Superficial Peritoneal Endometriosis (SPE)	Superficial implants, powder-burn lesions, red vesicles	Pelvic peritoneum, cul-de-sac	Superficial, non-infiltrating	Dysmenorrhea, mild pelvic pain, often asymptomatic

Genetic Architecture of Endometriosis Subphenotypes

Genome-Wide Association Studies (GWAS) Findings

Large-scale genetic studies have revealed significant differences in the genetic architecture of endometriosis subphenotypes. A landmark GWAS meta-analysis comprising 60,674 cases and 701,926 controls of European and East Asian ancestry identified 42 genome-wide significant loci comprising 49 distinct association signals [15]. Critically, this study demonstrated that ovarian endometriosis has a different genetic basis than superficial peritoneal disease, with distinct risk loci and effect sizes [15]. The identified signals explain up to 5.01% of disease variance, a threefold increase from previous studies, highlighting the importance of subphenotype stratification in genetic analyses.

The genetic heritability of endometriosis is estimated at approximately 50%, with common genetic variation accounting for 26% of cases [15]. Key implicated genes include WNT4, VEZT, GREB1, FN1, CCDC170, SYNE1, and ESR1, which play roles in sex hormone signaling, cell adhesion, proliferation, and inflammation [1] [15]. Deep infiltrating endometriosis demonstrates stronger genetic correlations with pain-related conditions including migraine, back pain, and multi-site pain, suggesting genetic contributions to central nervous system sensitization in chronic pain development [15].

Table 2: Selected Genetic Loci Associated with Endometriosis Subphenotypes

Gene/Locus	Reported Function	Ovarian Endometriosis	Deep Infiltrating Endometriosis	Superficial Peritoneal Endometriosis
WNT4	Sex development, estrogen signaling	Strong association	Moderate association	Weak association
VEZT	Cell adhesion	Strong association	Strong association	Moderate association
GREB1	Estrogen-regulated growth	Strong association	Moderate association	Weak association
FN1	Extracellular matrix organization	Moderate association	Strong association	Limited data
ESR1	Estrogen receptor	Moderate association	Strong association	Moderate association
CCDC170	Nuclear envelope organization	Strong association	Limited data	Limited data

Hormonal and Inflammatory Biomarkers

Subphenotype-specific biomarker profiles reflect underlying genetic differences. Aromatase (CYP19A1) shows increased expression in endometriotic tissues with a diagnostic sensitivity of 79% and specificity of 89% [1]. Progesterone resistance, characterized by reduced progesterone receptor-B (PR-B) expression and disrupted signaling, is particularly prominent in deep infiltrating lesions [1] [13]. Inflammatory biomarkers including macrophage migration inhibitory factor (MIF), interleukin-1 (IL-1), MMP-1, MMP-2, and MMP-3 demonstrate subphenotype-specific expression patterns, with elevated levels in deep infiltrating lesions contributing to tissue remodeling and invasion [1] [16].

Matrix metalloproteinases (MMPs) show distinct activity across subphenotypes, with pro-MMP-2 activity significantly higher in endometriotic lesions compared to eutopic endometrium and control tissue [16]. MMP-1 and MMP-3 protein levels are similarly elevated in lesions, creating a tissue microenvironment conducive to ectopic implantation and lesion establishment through extracellular matrix remodeling [16].

Experimental Protocols for Subphenotype-Specific Genetic Analysis

Protocol 1: GWAS Meta-Analysis for Subphenotype Stratification

Objective: To identify genetic variants associated with specific endometriosis subphenotypes through large-scale GWAS meta-analysis.

Materials:

• Genotype data from cases and controls with detailed subphenotype annotation
• High-performance computing infrastructure
• Quality control (QC) pipelines for genetic data
• GWAS analysis software (PLINK, SAIGE, REGENIE)

Methodology:

• Sample Collection and Phenotyping: Recruit endometriosis cases with surgical confirmation and detailed subphenotype characterization (ovarian, DIE, SPE). Include age-matched controls without endometriosis.
• Genotyping and Quality Control: Perform genome-wide genotyping using array technologies. Apply standard QC filters: call rate >98%, Hardy-Weinberg equilibrium p>1×10⁻⁶, minor allele frequency >1%. Exclude samples with sex discrepancies, excessive heterozygosity, or cryptic relatedness.
• Imputation: Use reference panels (1000 Genomes, HRC) for genotype imputation to increase variant coverage. Apply post-imputation QC (info score >0.8).
• Association Analysis: Conduct GWAS for each subphenotype separately using logistic regression adjusted for principal components. Apply genomic control to correct for residual population stratification.
• Meta-Analysis: Combine results across cohorts using fixed-effects or random-effects models. Apply heterogeneity testing (Cochran's Q) to identify subphenotype-specific effects.
• Variant Annotation: Annotate significant loci (p<5×10⁻⁸) with functional genomic data (eQTLs, chromatin states, protein interactions) to prioritize candidate genes.

Expected Outcomes: Identification of subphenotype-specific risk loci, calculation of subtype-specific heritability, and genetic correlation analyses between subphenotypes and related traits.

Protocol 2: Polygenic Risk Score Development for Endometriosis Subphenotypes

Objective: To construct and validate subphenotype-specific polygenic risk scores for endometriosis classification and risk prediction.

Materials:

• GWAS summary statistics for endometriosis subphenotypes
• Independent target dataset with genotype and phenotype data
• PRS calculation software (PRSice2, LDPred2, PRS-CS)
• Functional annotation databases (ENCODE, Roadmap Epigenomics)

Methodology:

• Base Data Preparation: Obtain GWAS summary statistics for each endometriosis subphenotype. Apply quality control: remove ambiguous SNPs, ensure consistent allele coding, filter for INFO score >0.9.
• Target Data Processing: Process independent target genotype data through standard QC pipeline. Calculate principal components to account for population structure.
• PRS Method Selection: Compare multiple PRS approaches: (1) clumping and thresholding, (2) Bayesian shrinkage methods (LDPred2), (3) continuous shrinkage priors (PRS-CS), (4) functional annotation-informed methods.
• PRS Calculation: Generate scores for each individual in the target dataset using optimized parameters from method comparison.
• Validation and Calibration: Assess PRS performance using regression models with subphenotype status as outcome. Calculate measures of discrimination (AUC-ROC, R²) and calibration. Perform cross-validation to avoid overfitting.
• Clinical Utility Assessment: Evaluate reclassification metrics (Net Reclassification Improvement) when adding PRS to clinical predictors. Establish risk thresholds for clinical application.

Expected Outcomes: Subphenotype-specific PRS with improved predictive accuracy compared to general endometriosis PRS, assessment of clinical utility for risk stratification and early intervention.

Protocol 3: Multi-omics Integration for Subphenotype Characterization

Objective: To integrate genomic, transcriptomic, and epigenomic data for comprehensive molecular characterization of endometriosis subphenotypes.

Materials:

• Endometriosis lesion tissues (ovarian, DIE, SPE) and matched eutopic endometrium
• DNA/RNA extraction kits
• Sequencing platforms (whole genome, RNA-seq, ATAC-seq)
• Multi-omics integration computational pipelines

Methodology:

• Sample Collection: Obtain surgical specimens of endometriosis subphenotypes with detailed clinical annotation. Include matched eutopic endometrium and control endometrium.
• DNA/RNA Extraction: Isolve high-quality DNA and RNA from frozen tissues. Assess quality (RIN >7.0 for RNA, DIN >7.0 for DNA).
• Sequencing: Perform whole genome sequencing, RNA sequencing, and ATAC sequencing on matched samples.
• Data Processing: Align sequences to reference genome. Call genetic variants, quantify gene expression, identify open chromatin regions.
• Integrative Analysis: Conduct molecular QTL analysis (eQTL, meQTL, caQTL) to link genetic variants to molecular phenotypes. Identify subphenotype-specific regulatory networks.
• Pathway Analysis: Perform gene set enrichment analysis to identify biological pathways specific to each subphenotype.

Expected Outcomes: Comprehensive molecular maps of endometriosis subphenotypes, identification of subtype-specific regulatory mechanisms, and functional validation of GWAS loci.

Signaling Pathways in Endometriosis Subphenotypes

The pathophysiology of endometriosis subphenotypes involves dysregulation of multiple signaling pathways. Ovarian endometriosis demonstrates prominent abnormalities in estrogen biosynthesis with overexpression of aromatase (CYP19A1) and steroidogenic factor-1 (SF-1) [1]. Deep infiltrating endometriosis shows activation of invasion-promoting pathways including MMP-mediated extracellular matrix degradation, epithelial-mesenchymal transition, and neuroangiogenesis [16]. Progesterone resistance, characterized by reduced PR-B expression and altered FKBP4 signaling, is common across subphenotypes but most pronounced in deep infiltrating disease [1] [13].

The Scientist's Toolkit: Research Reagent Solutions

Table 3: Essential Research Reagents for Endometriosis Subphenotype Studies

Reagent/Category	Specific Examples	Function/Application	Subphenotype Relevance
Genotyping Arrays	Global Screening Array, UK Biobank Axiom Array	Genome-wide variant genotyping	All subphenotypes - genetic association studies
Sequencing Kits	Illumina NovaSeq, PacBio HiFi, Oxford Nanopore	Whole genome, transcriptome, epigenome sequencing	All subphenotypes - comprehensive molecular profiling
Antibodies for IHC	Anti-aromatase (CYP19A1), anti-PR-B, anti-MMP-2, anti-CD56	Protein localization and quantification in tissues	Subphenotype-specific protein expression validation
Cell Culture Media	Stromal cell media, epithelial organoid culture systems	In vitro modeling of endometriosis lesions	Subphenotype-specific cellular behavior studies
Cytokine Assays	Luminex multiplex panels, ELISA kits for IL-1, MIF, IL-6	Quantification of inflammatory biomarkers	Subphenotype-specific inflammatory microenvironment
DNA/RNA Extraction Kits	QIAamp DNA FFPE, RNeasy, MagMAX for blood	Nucleic acid isolation from various sample types	Multi-omics analyses across subphenotypes
qPCR Reagents	TaqMan assays, SYBR Green master mixes	Gene expression validation	Candidate gene verification in subphenotypes
Methylation Arrays	Infinium MethylationEPIC	Genome-wide DNA methylation profiling	Epigenetic regulation in subphenotypes

Subphenotype-specific genetic analysis represents a paradigm shift in endometriosis research, moving beyond the traditional one-size-fits-all approach. The distinct genetic architectures of ovarian, deep infiltrating, and superficial peritoneal endometriosis underscore the necessity for stratified approaches in both basic research and clinical translation. Future directions should focus on: (1) expanding diverse ancestral representation in genetic studies, (2) integrating multi-omics data to functionalize genetic associations, (3) developing refined PRS with improved predictive accuracy across subphenotypes, and (4) translating genetic findings into subtype-specific therapeutic strategies. These advances will ultimately enable precision medicine approaches to endometriosis diagnosis, prevention, and treatment.

Endometriosis (EM) and Adenomyosis (AM) are prevalent gynecological disorders that pose significant diagnostic and therapeutic challenges in clinical practice. While both conditions share common symptoms, including chronic pelvic pain and infertility, they are recognized as distinct pathological entities. Endometriosis is characterized by the presence of endometrial-like tissue outside the uterine cavity, whereas adenomyosis involves the invasion of endometrial tissue into the myometrium.

Understanding the genetic architecture of these conditions is crucial for developing precise diagnostic tools and targeted therapies. This application note explores the fundamental genetic distinctions between endometriosis and adenomyosis, with a specific focus on implications for polygenic risk score (PRS) development for endometriosis subphenotypes. We present comprehensive genetic association data, detailed experimental protocols for analysis, and visualization of key biological pathways to advance research in this field.

Genetic Architecture and Distinctions

Key Genetic Findings from Recent Large-Scale Studies

Recent advances in genetic research have revealed substantial differences in the genetic architecture of endometriosis and adenomyosis. A landmark multi-ancestry genome-wide association study (GWAS) of approximately 1.4 million women, including 105,869 cases, identified 80 genome-wide significant associations, with 37 representing novel discoveries [17]. Crucially, this study identified five loci representing the first genetic variants ever reported for adenomyosis, providing initial insights into its unique genetic underpinnings [17].

Table 1: Summary of Key Genetic Associations for Endometriosis and Adenomyosis

Genetic Feature	Endometriosis	Adenomyosis
Number of GWAS loci	80 (37 novel) in recent large study [17]	5 first-ever variants reported [17]
Heritability	47-51% [9]	Not well established
PRS performance	OR = 1.57-1.59 per SD increase [11]	Not associated with endometriosis PRS [11]
Key pathways	Immune regulation, tissue remodeling, cell differentiation [17]	Shared and distinct mechanisms from endometriosis [18]
Multi-omics integration	Transcriptomic, epigenetic, and proteomic regulation across tissues [17]	Limited data available

Combinatorial analytics applied to UK Biobank and All of Us datasets have further elucidated these distinctions, revealing distinct mechanistic drivers for each condition, including multiple genes shared across both diseases and dozens of novel adenomyosis-associated genes not previously reported in endometriosis GWAS [18] [19]. This research supports the development of non-invasive differential diagnostic tools to improve patient triage across overlapping pelvic pain conditions [19].

Functional Characterization of Genetic Variants

The functional impact of endometriosis-associated genetic variants exhibits notable tissue specificity. Analysis of 465 endometriosis-associated GWAS variants using GTEx v8 database revealed that regulatory effects differ significantly across tissues [7]. In reproductive tissues (ovary, uterus, vagina), endometriosis-associated variants predominantly regulate genes involved in hormonal response, tissue remodeling, and cell adhesion [7]. In contrast, in peripheral blood and intestinal tissues, these variants primarily influence immune and epithelial signaling genes [7].

Key regulators such as MICB, CLDN23, and GATA4 have been consistently linked to hallmark pathways including immune evasion, angiogenesis, and proliferative signaling in endometriosis [7]. The tissue-specific regulatory patterns of these variants provide crucial insights for understanding the pathophysiology of endometriosis and its distinction from adenomyosis.

Experimental Protocols for Genetic Analysis

Genome-Wide Association Study Protocol

Objective: To identify genetic variants associated with endometriosis and adenomyosis risk across diverse ancestries.

Materials:

• Genotyping arrays: Illumina or Affymetrix platforms for genome-wide SNP coverage
• Bioinformatics tools: PLINK, METAL, GCTB for statistical analysis
• Cohort data: UK Biobank, All of Us, FinnGen, International Endogene Consortium data

Procedure:

• Sample Preparation and Quality Control
  • Extract DNA from participant blood samples using standard kits
  • Perform genotyping using selected platform
  • Apply quality control filters: call rate >98%, minor allele frequency >1%, Hardy-Weinberg equilibrium p > 1×10⁻⁶

• Association Analysis

  • Conduct logistic regression for case-control status using PLINK
  • Adjust for covariates: age, principal components, study-specific factors
  • Apply genomic control to correct for population stratification

• Meta-Analysis

  • Combine summary statistics across cohorts using METAL
  • Apply fixed-effects or random-effects models based on heterogeneity
  • Annotate significant variants with functional consequences using Ensembl VEP

• Fine-Mapping and Colocalization

  • Identify causal variants within associated loci using statistical fine-mapping
  • Perform colocalization analysis with eQTL datasets to identify target genes

Validation: Replicate findings in independent cohorts; perform functional validation through in vitro and in vivo models.

Polygenic Risk Score Development Protocol

Objective: To construct and validate polygenic risk scores for endometriosis subphenotypes.

Materials:

• GWAS summary statistics: From large meta-analyses
• Genotyping data: Target cohort with phenotype information
• Software: PRSice, plink, SBayesR, LDpred

Procedure:

• Clumping and Thresholding
  • Prune SNPs in linkage disequilibrium (r² < 0.1 within 250kb window)
  • Calculate PRS at multiple p-value thresholds (e.g., 0.001, 0.05, 0.1, 0.5, 1)
  • Select optimal threshold based on maximum R²

• Bayesian Polygenic Scoring

  • Apply SBayesR with default settings to adjust effect sizes
  • Exclude MHC region due to complex linkage structure
  • Calculate posterior effect sizes accounting for linkage disequilibrium

• PRS Calculation

  • Generate scores using plink1.9's score function: $PRS = \sum{i=1}^{n} βi × G_i$
  • Where βi is effect size of SNP i, Gi is genotype dosage

• Validation

  • Test association between PRS and disease status in independent cohorts
  • Assess discriminative accuracy using Area Under Curve (AUC) statistics
  • Evaluate stratification across subphenotypes (ovarian, peritoneal, infiltrating)

Application: The endometriosis PRS demonstrates significant association with all disease subtypes (ovarian OR = 1.72, infiltrating OR = 1.66, peritoneal OR = 1.51) but shows no association with adenomyosis, supporting distinct genetic architectures [11].

Signaling Pathways and Biological Mechanisms

Multi-Omic Integration in Endometriosis Pathogenesis

Genetic research has revealed that endometriosis risk variants exert their effects through transcriptomic, epigenetic, and proteomic regulation across multiple tissues [17]. These mechanisms converge on pathways involved in immune regulation, tissue remodeling, and cell differentiation [17].

Table 2: Key Pathways and Biological Processes in Endometriosis and Adenomyosis

Pathway Category	Specific Pathways	Implications
Immune regulation	Antigen processing and presentation, cytokine signaling	Altered immune surveillance, chronic inflammation [17] [7]
Tissue remodeling	Extracellular matrix organization, angiogenesis	Lesion establishment and growth [17]
Cell differentiation	Epithelial-mesenchymal transition, stem cell pathways	Tissue plasticity, invasive potential [17]
Metabolic pathways	Linoleic acid metabolism, glycerophospholipid metabolism	Shared alterations in EM and AM [20]
Hormone response	Estrogen receptor signaling, progesterone resistance	Hormone dependency of lesions [7]

Multi-omics studies integrating metabolomic and microbiome profiling have identified distinct metabolic and microbial signatures in both conditions. Specific pathways, including linoleic acid metabolism and glycerophospholipid metabolism, show significant alterations in both endometriosis and adenomyosis [20]. Notably, metabolites such as phosphatidylcholine 40:8 [PC(40:8)] exhibit marked changes in both conditions, suggesting some shared pathological features despite distinct genetic architectures [20].

The following diagram illustrates the integrated multi-omics approach to understanding endometriosis pathogenesis:

Diagram 1: Multi-omics integration in endometriosis research. This workflow illustrates how different data types inform our understanding of biological processes and clinical applications.

Hormonal Pathways and Their Genetic Regulators

A significant finding from PRS phenome-wide association studies is the association between genetic liability to endometriosis and lower testosterone levels, with Mendelian randomization analyses suggesting that lower testosterone may be causal for both endometriosis and clear cell ovarian cancer [9]. This highlights the importance of hormonal pathways in endometriosis pathogenesis and the potential for endocrine-focused interventions.

The tissue-specific regulatory patterns of endometriosis-associated variants further emphasize the role of hormonal responses. In reproductive tissues, these variants predominantly regulate genes involved in hormonal response, creating a permissive environment for lesion establishment and growth [7].

The Scientist's Toolkit: Research Reagent Solutions

Table 3: Essential Research Reagents for Genetic Studies of Endometriosis and Adenomyosis

Reagent/Category	Specific Examples	Application and Function
Genotyping platforms	Illumina Global Screening Array, Affymetrix Axiom Biobank array	Genome-wide SNP genotyping for association studies
Bioinformatics tools	PLINK, METAL, GCTB, PRSice, SBayesR	Statistical genetics analysis, meta-analysis, PRS calculation
eQTL resources	GTEx v8 database, eQTLGen Consortium	Mapping genetic variants to gene expression regulation
Cohort data	UK Biobank, All of Us, FinnGen, International Endogene Consortium	Large-scale genetic and phenotypic data for discovery and validation
Metabolomics platforms	Untargeted LC-MS (Liquid Chromatography-Mass Spectrometry)	Comprehensive metabolic profiling of endometrial samples [20]
Microbiome analysis	16S rRNA sequencing (5R approach)	Characterization of endometrial microbial communities [20]
Functional validation	CRISPR/Cas9 systems, organoid cultures, animal models	Mechanistic validation of genetic findings

The genetic distinctions between endometriosis and adenomyosis are becoming increasingly clear through large-scale genetic studies and multi-omics approaches. While they share some clinical manifestations and pathological features, their genetic architectures demonstrate significant differences, with unique risk loci and distinct regulatory mechanisms. These findings have profound implications for the development of polygenic risk scores specifically for endometriosis subphenotypes.

The experimental protocols and analytical frameworks presented in this application note provide researchers with robust methodologies for advancing this field. Future research directions should include expanded trans-ancestry genetic studies, functional characterization of novel loci, and the integration of polygenic risk scores with clinical factors for improved diagnosis and personalized treatment strategies.

Endometriosis, a chronic inflammatory and estrogen-dependent condition, affects approximately 10% of women of reproductive age and is a leading cause of pelvic pain and infertility [13] [1]. The diagnostic journey for patients is often protracted, spanning 7 to 12 years from symptom onset, largely due to the invasive nature of the current diagnostic gold standard—laparoscopic surgery with histological confirmation [9] [1]. This substantial delay underscores the critical need for non-invasive diagnostic strategies and improved risk stratification tools. In this context, the development of polygenic risk scores (PRS) for endometriosis subphenotypes represents a promising frontier. A PRS aggregates the effects of numerous genetic variants, each with small effect sizes, into a single quantitative measure of an individual's genetic liability to a disease [21]. Research confirms that a PRS for endometriosis captures an increased risk for the condition and its major subtypes, including ovarian, infiltrating, and peritoneal disease [21]. This application note details how the integration of hormonal and inflammatory pathway biology is fundamental to refining these genetic tools, thereby offering insights for researchers and drug development professionals aiming to deconstruct the heterogeneity of endometriosis and develop targeted therapeutic and diagnostic solutions.

Hormonal Dysregulation in Endometriosis

Key Hormonal Pathways

The hormonal landscape of endometriosis is characterized by two defining features: local estrogen dominance and progesterone resistance. Contrary to systemic circulation, local estrogen bioavailability is heightened within endometriotic lesions. This is driven by the overexpression of the enzyme aromatase (CYP19A1), which converts androgens into estrogens, and the downregulation of 17β-hydroxysteroid dehydrogenase type 2 (17β-HSD2), which inactivates estradiol [13] [1]. This creates a self-sustaining, estrogen-rich microenvironment. Concurrently, progesterone resistance—a failure of target tissues to respond adequately to progesterone—perpetuates lesion survival. This resistance is marked by a significant reduction in the progesterone receptor-B (PR-B) isoform, attributed to promoter hypermethylation and microRNA dysregulation [13].

A pivotal recent discovery from a polygenic risk score phenome-wide association study (PRS-PheWAS) is the genetic association between a higher liability to endometriosis and lower testosterone levels [9]. Follow-up Mendelian randomization analysis suggested that lower testosterone may have a causal effect on endometriosis risk, revealing a previously underappreciated role for androgen signaling in disease etiology [9].

Experimental Protocols for Hormonal Pathway Analysis

Protocol 1: Assessing Local Estrogen Biosynthesis in Eutopic Endometrium

• Objective: To quantify the expression of key enzymes in the estrogen activation pathway in menstrual blood or endometrial biopsy samples.
• Materials:
  • Sample Type: Menstrual blood or eutopic endometrial biopsy.
  • Key Reagents: Primers for CYP19A1 (aromatase), HSD17B2, and a reference gene (e.g., GAPDH); RNA extraction kit; reverse transcription and quantitative PCR (RT-qPCR) reagents.
  • Equipment: Real-time PCR system.
• Methodology:
  • Collect samples from confirmed endometriosis patients and healthy controls.
  • Extract total RNA and synthesize cDNA.
  • Perform RT-qPCR to measure the mRNA expression levels of CYP19A1 and HSD17B2.
  • Normalize expression to the reference gene and analyze using the ∆∆Ct method.
• Data Interpretation: A high CYP19A1 to HSD17B2 expression ratio is a strong indicator of local estrogen dominance. One study reported that aromatase expression in menstrual blood achieved an Area Under the Curve (AUC) of 0.977 for discriminating endometriosis patients from controls [1].

Protocol 2: Evaluating Progesterone Resistance via PR-B Immunohistochemistry

• Objective: To determine the protein expression and cellular localization of PR-B in endometrial stromal cells.
• Materials:
  • Sample Type: Formalin-fixed, paraffin-embedded (FFPE) endometrial tissue sections.
  • Key Reagents: Validated anti-PR-B antibody, immunohistochemistry (IHC) detection kit, and hematoxylin counterstain.
• Methodology:
  • Section FFPE tissue blocks and mount on slides.
  • Perform antigen retrieval and incubate with the primary anti-PR-B antibody.
  • Detect binding using a compatible IHC detection system and visualize with a chromogen.
  • Counterstain with hematoxylin to identify nuclei.
  • Score the staining intensity and proportion of PR-B positive stromal cells in a blinded manner.
• Data Interpretation: A significant loss of PR-B staining in the stromal compartment of the eutopic endometrium is a hallmark of progesterone resistance and is associated with impaired decidualization and infertility in endometriosis patients [13].

Table 1: Key Hormonal Biomarkers in Endometriosis

Biomarker	Molecular Function	Alteration in Endometriosis	Potential Diagnostic Utility
Aromatase (CYP19A1)	Converts androgens to estrogens	Overexpressed in lesions	High diagnostic accuracy (Sens: 79%, Spec: 89%) in meta-analysis [1]
Progesterone Receptor B (PR-B)	Mediates progesterone signaling	Significantly reduced in lesions	Indicator of progesterone resistance; correlates with infertility [13]
Testosterone	Androgen hormone	Genetically correlated with lower levels	Mendelian randomization suggests a causal, protective role [9]
17β-HSD2	Inactivates estradiol	Downregulated in lesions	Contributes to local estrogen dominance [13]
Nicotinamide N-methyltransferase (NNMT)	Modulates cell proliferation	Overexpressed, induced by estrogen	Potential new therapeutic target [1]

Inflammation and Immune Dysfunction

Chronic Inflammation and Altered Immunity

Endometriosis is a state of pervasive immune dysfunction and chronic inflammation. The peritoneal fluid of affected women becomes a pro-inflammatory milieu, characterized by altered populations and functions of immune cells [13]. Key alterations include:

• Macrophage Polarization: Macrophages, which constitute over 50% of immune cells in the peritoneal fluid, shift toward a "pro-endometriosis" phenotype. This is driven by neuroimmune communication, such as via calcitonin gene-related peptide (CGRP), leading to impaired clearance of ectopic cells and enhanced secretion of growth and angiogenic factors [13].
• Compromised Cytotoxicity: The cytotoxic activity of Natural Killer (NK) cells is severely blunted in both peripheral blood and peritoneal fluid, allowing ectopic endometrial cells to evade immune surveillance [13] [1].
• Cytokine Dysregulation: A complex interplay of pro-inflammatory cytokines, including IL-1, IL-6, and macrophage migration inhibitory factor (MIF), promotes angiogenesis, pain sensitization, and lesion establishment and growth [1].

This inflammatory state is not isolated but is genetically intertwined with broader autoimmune conditions. A recent study demonstrated significant genetic correlations between endometriosis and several immunological diseases, including rheumatoid arthritis (rg = 0.27), osteoarthritis (rg = 0.28), and multiple sclerosis (rg = 0.09). Mendelian randomization further suggested a potential causal relationship from endometriosis to rheumatoid arthritis (OR = 1.16) [22].

Experimental Protocols for Immune Profiling

Protocol 3: Flow Cytometric Analysis of Peritoneal Immune Cell Populations

• Objective: To characterize the composition and activation status of immune cells in the peritoneal fluid of endometriosis patients.
• Materials:
  • Sample Type: Peritoneal fluid aspirated during laparoscopy.
  • Key Reagents: Fluorescently conjugated antibodies against CD14 (macrophages), CD56/CD16 (NK cells), CD3 (T cells), CD4, CD8, and activation markers (e.g., CD69); flow cytometry staining buffer.
  • Equipment: Flow cytometer.
• Methodology:
  • Collect peritoneal fluid and separate cells by density gradient centrifugation.
  • Count cells and aliquot for staining.
  • Incubate cells with predefined antibody panels for surface markers.
  • Fix cells and acquire data on a flow cytometer.
  • Analyze data using flow cytometry software to identify immune cell subsets and their relative frequencies and activation states.
• Data Interpretation: Expect to find an increased proportion of macrophages (CD14+), a decreased proportion of cytotoxic CD56dimCD16+ NK cells, and a shift in T-helper cell balance toward Th2 and Th17 profiles in endometriosis patients compared to controls [13].

Protocol 4: Cytokine Profiling in Serum or Peritoneal Fluid

• Objective: To quantify the levels of pro- and anti-inflammatory cytokines in the systemic circulation and local pelvic environment.
• Materials:
  • Sample Type: Serum or peritoneal fluid.
  • Key Reagents: Multiplex cytokine array kit (e.g., for IL-1β, IL-6, IL-10, TNF-α, MIF).
  • Equipment: Luminex or MSD plate reader.
• Methodology:
  • Prepare samples and standards according to the kit protocol.
  • Add samples to the pre-coated plate and incubate.
  • After incubation with detection antibodies and streptavidin-conjugate, read the plate on the appropriate analyzer.
  • Generate a standard curve and calculate cytokine concentrations for each sample.
• Data Interpretation: Elevated levels of pro-inflammatory cytokines like IL-1, IL-6, and MIF are associated with the severity of endometriosis and its associated pain symptoms [1].

Table 2: Key Inflammatory and Immune Biomarkers in Endometriosis

Biomarker / Cell Type	Function	Alteration in Endometriosis	Research/Cinical Implication
M1/M2 Macrophages	Phagocytosis, tissue repair, angiogenesis	M1 dominant in eutopic endometrium; M2 dominant in lesions [13]	Drives inflammation and supports lesion survival; therapeutic target
CD56dimCD16+ NK cells	Cytotoxic activity	Severely reduced cytotoxicity [13]	Enables immune escape of ectopic cells
Macrophage Migration Inhibitory Factor (MIF)	Regulates immune responses, angiogenesis	Upregulated [1]	Contributes to inflammation and estrogen production
M2 Macrophages / γδ T cells	Immunomodulation	Infiltration associated with disease [23]	Identified as key players in the shared pathogenesis of EMs and RIF [23]
Rheumatoid Arthritis (RA)	Systemic autoimmune disease	Genetically correlated (rg = 0.27) [22]	Suggests shared biological mechanisms and comorbidity risk

Integration with Polygenic Risk Score Development

The biological pathways of hormone metabolism and inflammation provide a functional context for the genetic variants incorporated into PRS. The integration of these multi-omics layers is crucial for moving beyond a general disease PRS to subphenotype-specific prediction.

Connecting Genetics to Biology: The genetic variants identified in GWAS for endometriosis are enriched in genes involved in sex steroid hormone signaling, inflammatory pathways, and oncogenesis [24]. For instance, a recent study identified 51 methylation quantitative trait loci (mQTLs)—genetic variants that regulate DNA methylation—that were also associated with endometriosis risk, highlighting candidate genes like GREB1 and KDR that contribute to disease risk through epigenetic mechanisms [24]. This functionally annotates GWAS hits and prioritizes them for inclusion in refined PRS models.

Informing Subphenotype Stratification: The distinct hormonal and inflammatory profiles of different disease manifestations (e.g., ovarian vs. deep infiltrating endometriosis) or comorbidities (e.g., infertility vs. pain) can be used to validate and refine subphenotype-specific PRS. For example, a PRS was shown to be associated with all major subtypes of endometriosis but not with adenomyosis, confirming that the latter is driven by different genetic risk variants [21]. Furthermore, multi-omics analysis has identified shared diagnostic genes (e.g., PDIA4 and PGBD5) and immune microenvironment alterations (involving M2 macrophages and γδ T cells) between endometriosis and recurrent implantation failure (RIF), offering a molecular basis for stratifying patients based on infertility risk [23].

Enhancing Predictive Power: While the discriminative accuracy of a 14-SNP PRS alone is not yet sufficient for standalone clinical use (OR = 1.28-1.59 per SD increase) [21], combining PRS with classical clinical risk factors, hormonal levels (e.g., testosterone), and inflammatory biomarkers represents a powerful strategy for developing urgently needed risk stratification tools [9] [21] [1].

The Scientist's Toolkit: Research Reagent Solutions

Table 3: Essential Research Reagents for Endometriosis Pathway Analysis

Item	Function/Application	Example Use Case
SBayesR Software	Bayesian method for adjusting GWAS summary statistics to calculate improved PRS weightings [9].	Generating polygenic risk scores with optimized effect size estimates for association studies.
Illumina Infinium MethylationEPIC BeadChip	Genome-wide DNA methylation profiling of over 850,000 sites [24].	Identifying differential methylation patterns associated with menstrual cycle phase or disease state.
Validated PR-B Antibody	Specific detection of the Progesterone Receptor-B isoform in tissue sections via IHC.	Confirming progesterone resistance in endometrial stromal cells.
Multiplex Cytokine Panel (Luminex/MSD)	Simultaneous quantification of multiple cytokine and chemokine proteins in biofluids.	Profiling the inflammatory milieu in serum or peritoneal fluid.
Fluorochrome-conjugated Antibody Panel (CD14, CD56, CD16, CD3)	Immunophenotyping of immune cells from peritoneal fluid or blood by flow cytometry.	Characterizing shifts in macrophage and NK cell populations.
Primer Assays for CYP19A1, HSD17B2	Quantitative measurement of gene expression via RT-qPCR.	Assessing local estrogen biosynthesis activity in tissue or menstrual blood.

Visualizing Key Pathways and Workflows

Hormonal Signaling and Immune Crosstalk in Endometriosis

Core Pathways in Endometriosis Pathogenesis

PRS-PheWAS Workflow for Pathway Discovery

PRS-PheWAS Workflow for Comorbidity Discovery

Building Better Predictive Models: PRS Construction and Implementation Frameworks

SNP Selection and Weighting Strategies for Endometriosis PRS

Polygenic risk scores (PRS) have emerged as a powerful tool for quantifying an individual's genetic susceptibility to complex diseases. For endometriosis, a condition with a significant heritable component estimated at 47-52%, PRS represents a promising approach for risk prediction and stratification [25] [8]. The development of accurate PRS for endometriosis requires careful consideration of single nucleotide polymorphism (SNP) selection and weighting strategies, particularly when addressing the challenge of disease subphenotypes. This application note details standardized protocols for constructing and validating endometriosis PRS, with emphasis on translating genetic discoveries into biologically and clinically relevant tools.

Background and Significance

Endometriosis affects approximately 10% of women of reproductive age and is characterized by the presence of endometrial-like tissue outside the uterine cavity [11]. The disease demonstrates substantial heterogeneity in clinical presentation, with different subtypes including ovarian, peritoneal, and infiltrating endometriosis [11]. Genome-wide association studies (GWAS) have identified numerous genetic loci associated with endometriosis risk, enabling the development of PRS that aggregate the effects of multiple variants into a single quantitative measure [8].

The genetic architecture of endometriosis is polygenic, with each variant contributing modestly to disease risk. Early GWAS identified 12 SNPs at 10 independent loci, while more recent studies have expanded this to 42 significant loci [25] [9]. These discoveries provide the foundation for PRS development, though careful methodological approaches are required to optimize their predictive power and clinical utility.

SNP Selection Methods

Genome-Wide Significance Thresholding

The most straightforward approach to SNP selection involves including variants that reach genome-wide significance (p < 5 × 10⁻⁸) in GWAS. This method was employed in several early endometriosis PRS studies, such as one utilizing 14 lead SNPs from a large-scale meta-analysis [11]. While this approach ensures the inclusion of robustly associated variants, it may exclude SNPs with smaller but genuine effects, potentially limiting predictive accuracy.

Clumping and Thresholding (C+T)

The clumping and thresholding method represents an evolution beyond simple significance thresholding. This iterative process selects the SNP with the lowest p-value in a genomic region, removes SNPs in linkage disequilibrium (LD) with it, and repeats this process across the genome [26]. This strategy was applied in a PRS-PheWAS study that revealed an association between endometriosis genetic liability and testosterone levels [9].

Bayesian Methods

Advanced Bayesian methods represent the current state-of-the-art in SNP selection for PRS construction. Methods such as PRS-CS and SBayesR utilize shrinkage priors to model the genetic architecture of complex traits, allowing for the inclusion of a larger number of SNPs while accounting for LD structure [26] [9]. These approaches have demonstrated superior performance in endometriosis PRS applications, particularly in cross-ancestry contexts [27].

Table 1: Comparison of SNP Selection Methods for Endometriosis PRS

Method	Key Features	Advantages	Limitations	Representative Applications
Genome-Wide Significance	Includes SNPs with p < 5 × 10⁻⁸	High specificity for true associations	Limited number of SNPs; may miss polygenic signal	14-SNP PRS for endometriosis subtypes [11]
Clumping + Thresholding	LD-based pruning with p-value thresholds	Reduces redundancy; computationally efficient	Performance depends on threshold selection	PRS-PheWAS of endometriosis comorbidities [9]
Bayesian Methods	Shrinkage priors accounting for LD	Improved prediction accuracy; handles large SNP sets	Computationally intensive; requires careful prior specification	Cross-ancestry PRS in multi-ancestry GWAS [27]

Weighting Strategies

Effect Size Weighting

The most common approach to SNP weighting in PRS construction utilizes effect sizes (beta coefficients or odds ratios) derived from GWAS summary statistics. Each risk allele is weighted by its estimated effect size, with the overall PRS calculated as the weighted sum of risk alleles across all included SNPs [11]. This method was validated in a study of surgically confirmed endometriosis cases, where each standard deviation increase in PRS was associated with an odds ratio of 1.57 for endometriosis diagnosis [11].

Bayesian Shrinkage

Bayesian shrinkage methods, such as those implemented in SBayesR, adjust SNP effect sizes based on prior assumptions about the genetic architecture of the trait [9]. These approaches help mitigate the winner's curse phenomenon and provide more accurate effect size estimates, particularly for SNPs with modest associations. In a recent PRS-PheWAS, this approach demonstrated significant associations between endometriosis PRS and multiple biomarkers, including testosterone levels [9].

PRS-PGx Methods for Drug Response Prediction

For pharmacogenomic applications, novel weighting strategies have been developed that simultaneously model both prognostic (main) and predictive (interaction) effects. The PRS-PGx-Bayes method employs a Bayesian framework to estimate posterior distributions for both effect types, enabling the construction of separate prognostic and predictive PRS [28]. This approach has shown superior performance in predicting drug response compared to traditional disease PRS methods.

Experimental Protocols

GWAS Meta-Analysis for SNP Discovery

Objective: Identify genetic variants associated with endometriosis risk for inclusion in PRS.

Materials:

• Genotype and phenotype data from multiple cohorts
• GWAS analysis software (e.g., PLINK, SAIGE)
• Meta-analysis software (e.g., METAL)

Procedure:

• Perform quality control on genotype data for each cohort separately
• Conduct GWAS for endometriosis case-control status in each cohort
• Apply genomic control to correct for population stratification
• Combine summary statistics using fixed-effects meta-analysis
• Identify lead SNPs through clumping to ensure independence
• Validate associations in independent replication cohorts

This protocol was successfully implemented in a cross-ancestry meta-analysis of ∼1.4 million women, identifying 80 genome-wide significant associations including 37 novel loci [27].

PRS Construction and Validation

Objective: Construct PRS for endometriosis and validate its predictive performance.

Materials:

• GWAS summary statistics from discovery meta-analysis
• Independent target cohort with genotype and phenotype data
• PRS software (e.g., PLINK, PRS-CS, LDpred)

Procedure:

• Clump SNPs to remove those in high LD (r² > 0.1 within 250kb window)
• Calculate PRS using the formula: PRS = Σ(βᵢ × Gᵢ), where βᵢ is the effect size and Gᵢ is the genotype dosage for SNP i
• Assess association between PRS and endometriosis status using logistic regression
• Evaluate discriminative accuracy using area under the ROC curve (AUC)
• Stratify analysis by endometriosis subtypes if sample size permits
• Validate findings in independent cohorts to ensure generalizability

This methodology was applied in a study of Danish and UK Biobank cohorts, demonstrating significant association between PRS and all endometriosis subtypes [11].

Figure 1: PRS Development and Validation Workflow. This diagram illustrates the standardized pipeline for polygenic risk score construction, from initial quality control to final validation and application.

Subphenotype Stratification Analysis

Objective: Evaluate whether PRS performance varies across endometriosis subphenotypes.

Materials:

• Clinical data on endometriosis subphenotypes (e.g., ovarian, peritoneal, infiltrating)
• Genotype data for cases and controls
• Statistical analysis software (e.g., R, Python)

Procedure:

• Classify endometriosis cases into subphenotypes based on surgical records or ICD-10 codes
• Calculate PRS for all individuals in the cohort
• Perform logistic regression analyses for each subphenotype separately
• Compare effect sizes and predictive accuracy across subphenotypes
• Test for heterogeneity in PRS associations across subphenotypes

This approach revealed that PRS was associated with all major subtypes of endometriosis, with the strongest association for ovarian endometriosis (OR = 1.72) [11].

Data Integration and Functional Annotation

Multi-Omics Integration

Advanced PRS applications increasingly integrate multiple layers of genomic information. This includes transcriptomic data to identify expression quantitative trait loci (eQTLs), epigenetic data to map regulatory elements, and proteomic data to elucidate downstream pathways [27]. In a recent multi-ancestry study, integration of multi-omics data revealed that genetic variation influences endometriosis risk through transcriptomic, epigenetic, and proteomic regulation across multiple tissues [27].

Cross-Ancestry PRS

A significant challenge in PRS development is ensuring transferability across diverse ancestral populations. Recent efforts have focused on developing cross-ancestry PRS frameworks that incorporate data from multiple population groups [27]. These approaches typically involve:

• Multi-ancestry GWAS meta-analysis to identify shared risk variants
• Ancestry-specific effect size estimation
• LD structure accounting for population differences
• Validation in independent cohorts of diverse ancestry

Table 2: Performance Metrics of Endometriosis PRS Across Studies

Study Cohort	Sample Size (Cases/Controls)	Number of SNPs in PRS	Odds Ratio per SD	p-value	Subtype-Specific Effects
Danish Surgical Cohort [11]	249/348	14	1.59	2.57×10⁻⁷	Ovarian: OR=1.72; Infiltrating: OR=1.66
Danish Twin Registry [11]	140/316	14	1.50	0.0001	Not reported
UK Biobank [11]	2,967/256,222	14	1.28	<2.2×10⁻¹⁶	Not reported
UK Biobank PRS-PheWAS [9]	188,221 females	SBayesR	N/A	N/A	Association with testosterone levels

The Scientist's Toolkit

Table 3: Essential Research Reagents and Resources for Endometriosis PRS Studies

Resource Category	Specific Tools/Platforms	Application in Endometriosis PRS
Genotyping Arrays	Illumina Global Screening Array [29]	Genome-wide SNP genotyping for PRS calculation
Imputation Resources	TOPMed Imputation Server [29], 1000 Genomes Project [25]	Inference of ungenotyped variants using reference panels
GWAS Software	PLINK [9] [29], METAL [9], SAIGE	Association analysis and meta-analysis
PRS Methods	PRS-CS [26], SBayesR [9], LDpred [26]	Polygenic risk score calculation with various weighting approaches
Biobanks	UK Biobank [11] [9], FinnGen [9], All of Us [27]	Large-scale cohorts for discovery and validation
Functional Annotation	ENCODE [8], GTEx, GWAS Catalog [29]	Biological interpretation of risk loci

Signaling Pathways and Biological Mechanisms

Figure 2: Biological Pathways Implicated by Endometriosis PRS. Genetic risk variants for endometriosis aggregate in key signaling pathways involved in disease pathogenesis, including immune/inflammatory responses, hormonal regulation, developmental processes, and cellular functions.

Quality Control and Standardization

Genotype Quality Control

Robust quality control procedures are essential for reliable PRS construction. Standard protocols include:

• Sample-level QC: Exclusion based on call rate (<95%), heterozygosity outliers, sex discrepancies, and relatedness
• Variant-level QC: Exclusion based on call rate (<95%), Hardy-Weinberg equilibrium (p < 1×10⁻⁵), and minor allele frequency (<1%) [29]
• Population stratification: Adjustment using principal components or genetic relatedness matrices

Phenotype Harmonization

Consistent phenotype definitions across cohorts are critical for PRS validation. The Endometriosis Phenome and Biobanking Harmonization Project (EPHect) has developed standardized protocols for endometriosis data collection, including surgical and clinical phenotypes [25]. Implementation of these standards enables more reliable cross-study comparisons and meta-analyses.

SNP selection and weighting strategies for endometriosis PRS have evolved significantly, from early approaches using a handful of genome-wide significant variants to contemporary methods incorporating thousands of SNPs with Bayesian shrinkage. The continued expansion of GWAS sample sizes, improved representation of diverse ancestries, and integration of multi-omics data will further enhance the precision and utility of endometriosis PRS. These advances hold promise for refining endometriosis subphenotype classification, elucidating biological mechanisms, and ultimately improving risk prediction and targeted interventions.

Bayesian Methods and Machine Learning Approaches for PRS Optimization

Endometriosis, a complex gynecological disorder affecting 5-10% of women of reproductive age, presents substantial diagnostic challenges, with average delays of 4-11 years from symptom onset to definitive surgical diagnosis [30]. The disease demonstrates strong heritability estimates of 47-51% from twin studies and 26% from common SNP-based heritability, highlighting the significant genetic component that makes it amenable to polygenic risk scoring approaches [30] [9]. Current diagnostic limitations, including the requirement for invasive laparoscopic confirmation and the heterogeneity of clinical presentations, have created an urgent need for improved risk stratification tools [11].

Polygenic risk scores (PRS) aggregate the effects of numerous genetic variants across the genome to quantify an individual's genetic predisposition to a trait or disease. In endometriosis research, PRS has emerged as a promising approach for identifying high-risk individuals, elucidating biological pathways, and potentially reducing diagnostic delays [11]. However, standard PRS methods face several limitations, including limited predictive power, sensitivity to genetic architecture, and challenges in modeling the complex genetic underpinnings of endometriosis subphenotypes.

Bayesian methods and machine learning approaches offer sophisticated solutions to these limitations by incorporating prior biological knowledge, accommodating complex genetic architectures, and integrating diverse data types. This application note provides detailed protocols and methodologies for implementing these advanced computational techniques to optimize PRS for endometriosis subphenotype research, specifically targeting researchers, scientists, and drug development professionals working in this field.

Bayesian Methods for PRS Optimization

Theoretical Foundations

Bayesian methods for PRS construction fundamentally differ from traditional approaches by incorporating prior distributions over SNP effect sizes, allowing for more flexible modeling of genetic architecture. The core Bayesian linear regression framework is expressed as:

y = Xβ + ε

Where y is the vector of phenotypic measurements, X is the genotype matrix, β is the vector of effect sizes, and ε captures residual effects [31]. The Bayesian approach specifies prior distributions on the effect sizes β, which are then updated through the likelihood to obtain posterior distributions given the observed data [32].

The key advantage of Bayesian methods lies in their ability to model genetic architectures through specific prior distributions. The spike-and-slab prior implements a mixture distribution:

βj ~ πN(βj; 0, σβ²) + (1 - π)δ0

This formulation specifies that each SNP effect size βj follows a normal distribution with probability π (the fraction of causal variants) or is exactly zero with probability (1-π) [31]. Continuous shrinkage priors, such as those implemented in PRS-CS, provide an alternative approach that allows for marker-specific adaptive shrinkage, eliminating the need for discrete mixture distributions while effectively modeling varying genetic architectures [32].

Implementation Protocols

VIPRS (Variational Inference of Polygenic Risk Scores)

VIPRS utilizes variational inference to approximate posterior distributions for effect sizes, offering computational advantages over traditional Markov Chain Monte Carlo (MCMC) methods [31].

Protocol Steps:

• Data Preparation: Process GWAS summary statistics and align with LD reference panel (e.g., 1000 Genomes European sample)
• Model Configuration:
  • Set spike-and-slab prior parameters
  • Initialize variational distributions
• Parameter Estimation:
  • Iteratively update variational parameters until convergence
  • Monitor evidence lower bound (ELBO) for convergence assessment
• Effect Size Extraction: Output posterior mean effect sizes for PRS calculation

Key Advantages: VIPRS demonstrates competitive predictive accuracy while being more than twice as fast as MCMC-based approaches, with robust performance across diverse genetic architectures [31].

PRS-CS and PRS-CS-auto

PRS-CS employs a continuous shrinkage prior that enables conjugate block updates of SNP effect sizes, providing accurate modeling of local LD patterns [32].

Protocol Steps:

• Input Preparation:
  • GWAS summary statistics
  • External LD reference panel (1000 Genomes recommended)
• Global Shrinkage Parameter Setting:
  • PRS-CS: Search fixed ϕ values, select optimal value via validation dataset
  • PRS-CS-auto: Place half-Cauchy prior ϕ¹/² ~ C⁺(0,1), automatically learned from data
• Gibbs Sampling: Implement block update of effect sizes within LD regions
• PRS Construction: Calculate scores using posterior mean effect sizes

Performance Characteristics: PRS-CS demonstrates substantial improvements in prediction accuracy across varying genetic architectures, particularly with large training sample sizes [32].

Workflow Visualization

Figure 1: Bayesian PRS Optimization Workflow

Machine Learning Approaches

Integrated Predictive Modeling

Machine learning approaches enable the integration of polygenic risk scores with diverse clinical and demographic variables to improve endometriosis prediction. The gradient boosting algorithm CatBoost has demonstrated particularly strong performance in this domain, achieving an area under the ROC curve (AUC) of 0.81 when combining genetic, clinical, and lifestyle factors [30].

Protocol: Integrated ML Pipeline for Endometriosis Risk Prediction

• Feature Engineering and Selection

  • Input Features:
    • PRS (19-72 SNV models) [33]
    • Clinical factors: irritable bowel syndrome, menstrual cycle length [30]
    • Demographic data: age, BMI [33]
    • Reproductive history: age at menarche, number of live births [9]
  • Missing Data Handling: Implement multiple imputation or missingness indicators
  • Feature Importance: Apply SHAP (SHapley Additive exPlanations) for model interpretability

• Model Training with CatBoost

  • Hyperparameter tuning via cross-validation
  • Regularization to prevent overfitting
  • Implementation of early stopping criteria

• Model Interpretation

  • Calculate feature importance scores
  • Generate individual-level SHAP values for clinical translation
  • Validate identified risk factors against existing clinical knowledge

PRS-PheWAS for Pleiotropy Analysis

Polygenic risk score phenome-wide association studies (PRS-PheWAS) enable the systematic investigation of genetic liability to endometriosis across diverse phenotypes and biomarkers, revealing important pleiotropic effects [9].

Protocol: PRS-PheWAS Implementation

• Cohort Definition

  • Primary analysis: Unrelated European females (n = 188,221)
  • Sensitivity analysis: Females without endometriosis diagnosis (n = 182,789)
  • Sex-specific analysis: Males (n = 159,855) to identify sex-specific pathways

• Phenotype Processing

  • Map ICD-10 codes to phecodes
  • Transform blood and urine biomarkers (log transformation)
  • Apply quality control to reproductive factors

• Association Testing

  • Logistic regression for binary traits (phecodes)
  • Linear regression for continuous biomarkers
  • Covariate adjustment: age, genetic principal components

Performance Comparison and Application to Endometriosis

Method Performance Metrics

Table 1: Performance Comparison of PRS Methods for Endometriosis

Method	Key Features	Genetic Architecture Handling	Computational Efficiency	Reported Performance (OR/AUC)
Traditional PRS	Clumping and thresholding	Limited	High	OR: 1.57-1.72 for subtypes [11]
VIPRS	Variational inference, spike-and-slab prior	Robust	High (2x faster than MCMC)	Competitive with state-of-art [31]
PRS-CS	Continuous shrinkage priors, LD modeling	Excellent	Moderate	Improved accuracy vs. alternatives [32]
CatBoost ML	Integrated PRS + clinical factors	N/A	Moderate	AUC: 0.81 [30]

Endometriosis Subphenotype Applications

Bayesian approaches have demonstrated particular utility in dissecting the genetic architecture of endometriosis subphenotypes. Research has shown that PRS can capture increased risk across all types of endometriosis rather than specific locations, with odds ratios of 1.72 for ovarian endometriosis, 1.66 for infiltrating endometriosis, and 1.51 for peritoneal endometriosis [11]. Furthermore, these approaches have revealed that endometriosis PRS is not associated with adenomyosis, suggesting distinct genetic etiologies despite clinical similarities [11].

The application of Bayesian methods to gene identification has successfully prioritized high-confidence candidate genes, with studies identifying 24 genes with high-confidence scores including HLA-DQB1 and PPARA as central to the endometriosis network [34]. These findings provide biological insights that may inform future therapeutic development.

Biomarker and Comorbidity Associations

PRS-PheWAS approaches have revealed significant associations between genetic liability to endometriosis and multiple biomarkers, most notably identifying an association with lower testosterone levels that may be causal for both endometriosis and clear cell ovarian cancer [9]. This finding highlights the value of these methods for uncovering novel biological pathways and potential therapeutic targets.

Table 2: Key Endometriosis PRS Associations from PRS-PheWAS

Category	Specific Association	Direction	Potential Clinical Relevance
Reproductive Factors	Menstrual cycle length	Positive	Early risk indicator
Comorbid Conditions	Irritable bowel syndrome	Positive	Diagnostic clarification
Biomarkers	Testosterone levels	Negative	Novel therapeutic target
* Psychiatric Comorbidities*	Depression	Positive	Comprehensive patient care

Research Reagent Solutions

Table 3: Essential Research Reagents and Computational Tools

Resource Category	Specific Tool/Dataset	Application	Key Features
GWAS Summary Statistics	Sapkota et al. 2017 meta-analysis [9]	PRS weight derivation	14,926 cases; 189,715 controls
LD Reference Panels	1000 Genomes European sample [32]	LD adjustment	N = 503; Population-specific
Biobank Data	UK Biobank [30] [9]	Method validation	~500,000 participants; Rich phenotyping
Software Tools	GCTB 2.02 (SBayesR) [9]	Bayesian PRS	Summary-based Bayesian analysis
Clinical Validation Cohorts	Western Danish endometriosis cohort [11]	Clinical translation	Surgically confirmed cases

Integrated Workflow for Endometriosis Subphenotype Research

Figure 2: Endometriosis Subphenotype Research Workflow

Bayesian methods and machine learning approaches significantly advance PRS optimization for endometriosis research by improving predictive accuracy, enabling subphenotype stratification, and uncovering novel biological insights. The integration of these computational approaches with comprehensive phenotypic data from biobanks provides a powerful framework for addressing the diagnostic challenges in endometriosis and facilitating the development of targeted therapeutic strategies. Future directions should focus on increasing ancestral diversity in genetic studies, refining subphenotype definitions, and translating these computational advances into clinical tools for risk stratification and early intervention.

The diagnostic odyssey for endometriosis, often protracted by 7 to 12 years, underscores the critical need for innovative risk stratification tools [1]. Current reliance on laparoscopic surgery for definitive diagnosis creates a significant barrier to timely intervention. Polygenic risk scores (PRS), which aggregate the effects of numerous genetic variants, offer a promising avenue for understanding disease susceptibility. However, the discriminative accuracy of PRS alone remains insufficient for standalone clinical prediction, with odds ratios (OR) for endometriosis typically ranging from 1.28 to 1.59 per standard deviation increase in PRS [11]. This protocol details methodologies for integrating PRS with non-genetic risk factors to enhance predictive power and provide a more comprehensive framework for endometriosis research and potential future clinical application.

Quantitative Evidence Base for Integration

Empirical evidence from large-scale biobank studies provides a solid foundation for integrating PRS with clinical risk factors. The interactions between genetic susceptibility and clinical manifestations are complex and multidimensional, as summarized in the table below.

Table 1: Evidence for PRS-Clinical Factor Interactions in Endometriosis

Factor Category	Specific Factor	Nature of Interaction with PRS	Key Findings	Source
Comorbidities	Uterine Fibroids, Heavy Menstrual Bleeding, Dysmenorrhea	Significant interaction	Absolute increase in endometriosis prevalence greater in individuals with high PRS vs. low PRS when comorbidity present.	[35] [36]
Comorbidity Burden	Overall diagnosed condition count	Negative correlation in cases	Comorbidity burden positively correlated with PRS in women without endometriosis but negatively correlated in diagnosed cases.	[35] [36]
Hormonal Biomarkers	Testosterone	Causal relationship suggested	Genetic liability to lower testosterone identified as potentially causal for endometriosis via Mendelian Randomisation.	[9]
Disease Subtypes	Ovarian, Infiltrating, Peritoneal	PRS association varies	PRS associated with all subtypes (ORs: Ovarian=1.72, Infiltrating=1.66, Peritoneal=1.51).	[11]
Clinical Presentation	Spread of disease, GI tract involvement	Inverse association	Higher PRS unexpectedly associated with less spread and fewer GI symptoms in one clinical cohort.	[37]

Protocol for PRS and Clinical Risk Factor Integration

Polygenic Risk Score Calculation

Objective: To generate a standardized endometriosis PRS for research applications.

Workflow Overview: The following diagram outlines the core PRS calculation workflow.

Methodology:

• GWAS Summary Statistics: Utilize the most recent, well-powered endometriosis GWAS meta-analysis. A recommended starting point is the meta-analysis of Sapkota et al. (2017) and FinnGen Release 8, yielding 14,926 cases and 189,715 controls for SNP effect size estimation [35] [9].
• Summary Statistics Processing:
  • Apply quality control: Remove duplicate SNPs (keeping the one with the lowest p-value), restrict to minor allele frequency (MAF) >1%, and exclude the major histocompatibility complex (MHC) region due to its complex linkage disequilibrium [35].
  • Refine SNP effect sizes using a Bayesian method such as SBayesR (implemented in GCTB version 2.02) to improve PRS prediction accuracy [35] [9].
• Target Cohort Genotyping:
  • Genotype research participants using a high-density array (e.g., Illumina Global Screening Array) [37].
  • Perform rigorous quality control: exclude samples with high missingness (>5%), remove related individuals (PI-HAT > 0.1875), and exclude SNPs with high missingness (>5%), low MAF (<1%), or deviation from Hardy-Weinberg equilibrium (p < 1×10⁻⁵) [37].
  • Impute genotypes to a reference panel (e.g., TOPMed) and filter for imputation quality (INFO score >0.80) [37].
• PRS Calculation:
  • Calculate the PRS using PLINK 1.9/2.0 --score function, applying the formula: PRS = Σ (β_i * G_ij) where βi is the effect size of SNP i and Gij is the allele count (0,1,2) for SNP i in individual j [35] [11].
  • Standardize the PRS to a Z-score within the cohort for subsequent analyses.

Ascertainment of Clinical Risk Factors and Comorbidities

Objective: To systematically collect and code non-genetic risk factors for integration with PRS.

Methodology:

• Data Sources:
  • Medical Records: Extract diagnoses from hospital and primary care records using the International Classification of Diseases, Tenth Revision (ICD-10) [35] [9]. Simplify 5-character codes to 4-character codes to reduce granularity. For endometriosis comorbidity searches, exclude codes related to injuries (Chapter XIX) and external causes (Chapter XX) [35].
  • Patient Questionnaires: Administer standardized instruments to capture symptoms not fully recorded in diagnostic codes, such as:
    • Pelvic pain and dysmenorrhea (painful periods) [35] [36].
    • Gastrointestinal symptoms using validated tools like the Visual Analog Scale for Irritable Bowel Syndrome (VAS-IBS) [37].
    • Reproductive history, including age at menarche, menstrual cycle characteristics, and infertility [9].
• Defining Index Comorbidities: Based on established associations, prioritize conditions such as uterine fibroids, heavy menstrual bleeding, dysmenorrhea, irritable bowel syndrome, diverticular disease, and asthma for interaction analyses [35] [36].

Integrated Risk Modeling and Statistical Analysis

Objective: To quantify the combined effect of PRS and clinical factors on endometriosis risk.

Methodology:

• Model Specification: Employ multivariable logistic regression to model the log-odds of endometriosis diagnosis.
  • Base Model: logit(P) = β₀ + β₁(PRS) + Σγ_i(Covariates_i) where covariates include age, genetic principal components (PCs 1-10), and other relevant confounders [35] [9].
  • Full Interactive Model: logit(P) = β₀ + β₁(PRS) + β₂(Clinical Factor) + β₃(PRS * Clinical Factor) + Σγ_i(Covariates_i) The coefficient β₃ tests for the interaction between genetic liability and the clinical factor.
• Analysis of Comorbidity Burden:
  • Calculate the total count of unique 3-character ICD-10 diagnoses for each participant, excluding those related to pregnancy and those with significantly decreased risk in endometriosis cases [35].
  • Correlate this comorbidity burden count with the PRS, stratifying by endometriosis case/control status.
• Model Performance Evaluation: Assess discriminatory power by calculating the Area Under the Receiver Operating Characteristic Curve (AUC). Compare the AUC of the model with PRS alone, clinical factors alone, and the combined model [11] [33].

The Scientist's Toolkit: Research Reagent Solutions

Table 2: Essential Reagents and Resources for Integrated PRS-Endometriosis Research

Item/Category	Specification/Example	Primary Function in Protocol
Genotyping Array	Illumina Global Screening Array	High-throughput genotyping of DNA samples to obtain raw SNP data.
Imputation Reference Panel	TOPMed (Version R2 on GRC38)	Statistical inference of non-genotyped markers to increase SNP density, using a large, diverse reference panel.
GWAS Summary Statistics	Sapkota et al. (2017) meta-analysis + FinnGen R8	Provides effect sizes (beta coefficients) for risk alleles used to weight the PRS.
Bioinformatics Software - QC	PLINK 1.9/2.0	Data management and quality control of genotype data (filtering, relatedness checks).
Bioinformatics Software - PRS	GCTB (for SBayesR), PLINK --score	Refines SNP weights and calculates the polygenic risk score for each individual.
Phenotype Codification	ICD-10 to Phecode Mapping (v1.2)	Aggregates specific ICD-10 codes into broader, clinically meaningful phenotypes for association testing.
Statistical Software	R Statistical Environment	Data analysis, statistical modeling (logistic regression), and visualization.

Advanced Integrative Concepts

Beyond SNP-based PRS: Methylation Risk Scores (MRS)

Emerging evidence suggests that integrating epigenetic markers can further improve risk prediction. A recent study developed an MRS for endometriosis using endometrial tissue methylation data from 908 samples [38].

• Variance Explained: DNA methylation captured 15.4% of the variance in endometriosis status, independent of common genetic variants [38].
• Predictive Performance: The best-performing MRS achieved an AUC of 0.67. When MRS was combined with PRS, the classification performance was consistently higher than with PRS alone, indicating that MRS captures non-genetic and environmental risk components [38].
• Protocol Consideration: For a comprehensive risk profile, researchers can develop an MRS from methylation data (e.g., from Illumina EPIC arrays) using methods like MLM-based omic association (MOA) and incorporate it as an additional predictor in integrated risk models [38].

Pathway and Pleiotropy Analysis

Understanding the biological pathways connecting genetic risk to comorbidities can illuminate disease mechanisms.

Key Pathway Insight: A PRS-phenome-wide association study (PheWAS) revealed an association between genetic liability for endometriosis and lower testosterone levels. Mendelian randomization analysis suggested that lower testosterone may have a causal effect on endometriosis risk [9]. This finding points to specific hormonal pathways that may be therapeutic targets and should be considered when integrating hormonal profiles into risk models.

Polygenic risk scores (PRS) have emerged as a valuable tool for quantifying an individual's genetic susceptibility to complex diseases like endometriosis. Traditionally, PRS are calculated from the cumulative effect of common single nucleotide polymorphisms (SNPs) identified through genome-wide association studies (GWAS) [11]. For endometriosis, these common variants explain only a portion of the known heritability, estimated at approximately 50% based on twin and family studies [39] [40]. This "missing heritability" problem has prompted increased investigation into the role of rare genetic variants and structural variations in endometriosis pathogenesis.

Endometriosis affects approximately 10% of reproductive-aged women globally and presents a substantial diagnostic challenge, with delays of 7-10 years from symptom onset to definitive diagnosis [41] [39]. The development of more accurate PRS that incorporate diverse genetic elements could significantly improve early detection and risk stratification, particularly for different endometriosis subphenotypes. This protocol outlines comprehensive methodologies for identifying, analyzing, and integrating rare variants and structural variations into endometriosis PRS frameworks.

Table 1: Types of Genetic Variations in Endometriosis

Variant Category	Definition	Detection Method	Contribution to Endometriosis
Common Variants	SNPs with minor allele frequency (MAF) >5%	GWAS	~26% of polygenic risk [29]
Rare Variants	SNPs with MAF <1%	Whole-exome sequencing, whole-genome sequencing	Increased risk in familial cases [2]
Copy Number Variants (CNVs)	Structural variations >1kb	High-density microarrays, sequencing	Rare, large-effect deletions [42]
Regulatory Variants	Non-coding variants affecting gene expression	eQTL analysis, functional genomics	Tissue-specific effects [39] [43]

Rare Variant Detection and Analysis

Whole-Exome Sequencing for Rare Coding Variants

Whole-exome sequencing (WES) provides an efficient approach for identifying rare coding variants with potentially significant functional impact in endometriosis patients.

Protocol: Family-Based WES Analysis

• Sample Selection: Prioritize multi-generational families with multiple affected individuals to enhance detection of co-segregating rare variants [2]. The study should include at least three affected family members across different generations.

• DNA Extraction and Library Preparation: Extract genomic DNA from peripheral blood leukocytes using standardized protocols. Prepare sequencing libraries with exome capture kits (e.g., Illumina Exome Panel) following manufacturer specifications.

• Sequencing Parameters: Sequence on platforms such as Illumina NovaSeq with a minimum coverage of 100x to ensure reliable variant calling. Include both affected and unaffected family members for comparison.

• Bioinformatic Analysis:

  • Align FASTQ files to reference genome (GRCh37/hg19 or GRCh38) using BWA-MEM
  • Perform duplicate removal and base quality recalibration
  • Call variants with tools such as FreeBayes or GATK HaplotypeCaller
  • Annotate variants using Ensembl VEP or similar tools

• Variant Filtering Strategy:

  • Focus on rare variants (MAF <0.01 in population databases)
  • Prioritize protein-altering variants (missense, frameshift, stop-gain/loss)
  • Identify variants co-segregating with affected status
  • Exclude common polymorphisms and technical artifacts

Table 2: Candidate Rare Variants Identified in Familial Endometriosis

Gene	Variant	Function	Evidence
LAMB4	c.3319G>A (p.Gly1107Arg)	Extracellular matrix protein	Co-segregation in multigenerational family [2]
EGFL6	c.1414G>A (p.Gly472Arg)	Angiogenesis regulator	Co-segregation in multigenerational family [2]
NAV3	Rare variants	Neuronal development	Potential role in pain perception [2]
NPSR1	High-penetrance variants	Neuropeptide signaling	Associated with endometriosis risk [2]

Functional Validation of Rare Variants

After identifying candidate rare variants, functional validation is essential to establish pathogenicity:

• In Silico Prediction: Utilize tools like SIFT, PolyPhen-2, and CADD to predict variant impact
• Expression Studies: Quantify gene expression in endometriosis lesions versus eutopic endometrium
• Functional Assays: Develop cellular models (e.g., CRISPR-edited cell lines) to assess variant effects on proliferation, invasion, and hormone response

Structural Variation Analysis

Copy Number Variant Detection

Copy number variants (CNVs) are structural variations ≥1kb in length that contribute significantly to genomic diversity and disease susceptibility.

Protocol: Genome-Wide CNV Analysis

• Sample Preparation and Quality Control:

  • Use high-density genotyping arrays (e.g., Illumina HumanOmniExpress) or whole-genome sequencing
  • Apply stringent quality filters: LRR-SD <0.3, B-allele frequency drift <0.01, call rate >98%
  • Include population-matched controls (recommended ratio: 1 case to 8-10 controls)

• CNV Calling and Filtering:

  • Process raw signal intensity data (Log R Ratio and B Allele Frequency) using PennCNV or similar algorithms
  • Apply a minimum window of 10 consecutive probes for CNV detection
  • Implement empirical filters to reduce false positives:
    • Remove CNVs in known genomic regions with high technical variability
    • Exclude CNVs with low confidence scores (<30)
    • Filter out common CNVs present in >1% of control population

• Statistical Analysis:

  • Compare global CNV burden (count, size, distribution) between cases and controls
  • Perform association testing at specific genomic loci
  • Adjust for multiple testing using Bonferroni correction or false discovery rate

• Technical Validation:

  • Confirm putative CNVs using alternative platforms (e.g., Affymetrix CytoScan HD)
  • Validate with quantitative PCR or droplet digital PCR

A recent CNV analysis in endometriosis identified three significant deletions associated with disease risk: a deletion at SGCZ on 8p22 (OR = 8.5, P = 7.3×10⁻⁴), a deletion in MALRD1 on 10p12.31 (OR = 14.1, P = 5.6×10⁻⁴), and a deletion at 11q14.1 (OR = 33.8, P = 5.7×10⁻⁴) [42]. These CNV loci were detected in 6.9% of affected women compared to 2.1% in the general population [42].

Integrating Diverse Variants into Polygenic Risk Models

Multi-Variant PRS Framework

Developing comprehensive PRS that incorporate both common and rare variants requires specialized statistical approaches:

• Variant Weighting:

  • Common variants: Weight by effect sizes from large GWAS meta-analyses
  • Rare variants: Apply burden tests or variance-component tests
  • CNVs: Incorporate as binary variables or dosage effects

• Integration Methods:

  • Develop separate scores for different variant classes then combine
  • Use machine learning approaches (random forests, gradient boosting) to integrate diverse features
  • Apply Bayesian methods to incorporate prior probabilities of pathogenicity

• Subphenotype Stratification:

  • Calculate subtype-specific scores for ovarian, peritoneal, and infiltrating endometriosis
  • Account for clinical presentation differences (pain symptoms, infertility)

Research demonstrates that PRS based on 14 common variants shows differential association with endometriosis subtypes: ovarian (OR = 1.72), infiltrating (OR = 1.66), and peritoneal (OR = 1.51) [11]. This suggests that incorporating additional variant types may further improve subphenotype discrimination.

Functional Annotation and Pathway Integration

Enhance PRS interpretation by incorporating functional genomic data:

• eQTL Analysis: Identify variants regulating gene expression in relevant tissues (uterus, ovary, immune cells) using resources like GTEx [43]
• Epigenetic Annotation: Overlap variants with endometriosis-relevant epigenetic marks (DNA methylation, histone modifications)
• Pathway Analysis: Group variants by biological pathways (sex steroid regulation, inflammation, cell adhesion)

Regulatory variants in genes such as IL-6, CNR1, and IDO1 have been identified through eQTL analysis and may interact with environmental factors like endocrine-disrupting chemicals [39].

Experimental Protocols

Comprehensive Endometriosis Genotyping Workflow

The following protocol outlines an integrated approach for detecting diverse variant types in endometriosis research studies.

Regulatory Variant Analysis Workflow

This protocol details the identification and functional characterization of regulatory variants in endometriosis.

The Scientist's Toolkit

Table 3: Essential Research Reagents and Platforms for Endometriosis Genetic Studies

Category	Item	Specifications	Application
Sequencing Platforms	Illumina NovaSeq	100x coverage, 150bp paired-end	Whole genome/exome sequencing [2]
Genotyping Arrays	Illumina HumanOmniExpress	~720,000 markers	CNV detection, common variants [42]
CNV Calling Software	PennCNV	Minimum 10 probes, LRR-SD filter	Structural variant identification [42]
Variant Annotation	Ensembl VEP	GRCh37/hg38, population frequencies	Functional consequence prediction [43]
eQTL Resources	GTEx Portal v8	Multiple tissues, FDR <0.05	Regulatory variant mapping [43]
Statistical Analysis	PLINK, PRSice	QC filters, clumping, weighting	Polygenic risk score calculation [11] [29]
Functional Validation	CRISPR-Cas9	Gene editing, reporter assays	Mechanistic studies of priority variants [39]

Incorporating rare genetic variants and structural variations into polygenic risk models represents a promising frontier in endometriosis research. The protocols outlined here provide a comprehensive framework for detecting, validating, and integrating these diverse genetic elements to enhance PRS accuracy and clinical utility. As research in this area advances, multi-ancestry studies and standardized bioinformatic pipelines will be essential for developing PRS that effectively stratify risk across diverse populations and endometriosis subphenotypes. This integrated approach ultimately promises to improve early detection, personalized treatment strategies, and our fundamental understanding of endometriosis pathogenesis.

Addressing Clinical Translation Challenges in Endometriosis PRS Development

The Area Under the Receiver Operating Characteristic Curve (AUC) serves as the most prevalent metric for evaluating the discriminative ability of polygenic risk models, quantifying how well a model distinguishes between individuals who will or will not develop a disease [44]. In the context of endometriosis subphenotype research, accurate prediction is particularly challenging due to the disease's heterogeneity, multifactorial etiology, and the subtle contribution of individual genetic variants [13] [1]. While AUC provides a valuable overview of model performance, reliance on this single metric presents significant limitations, especially when evaluating incremental improvements offered by polygenic risk scores (PRS) for complex traits.

Recent research highlights that AUC values often fail to detect clinically relevant improvements when new genetic markers are added to existing models, a phenomenon particularly problematic in endometriosis research where the goal is to stratify risk across diverse disease manifestations [44] [45]. This application note examines the inherent limitations of AUC, explores complementary metrics and methodological refinements, and provides structured experimental protocols to enhance the predictive power and clinical utility of PRS models for endometriosis subphenotypes.

Critical Limitations of AUC in Polygenic Risk Assessment

Conceptual and Statistical Shortcomings

The AUC metric possesses several inherent limitations that constrain its utility for comprehensive model evaluation:

• Insensitivity to Clinically Meaningful Improvements: The AUC can remain virtually unchanged even when risk predictions improve meaningfully for substantial portions of the population, particularly when models already demonstrate moderate discriminative ability (AUC > 0.70) [44]. This insensitivity stems from AUC's focus on ranking rather than the magnitude of risk differences.

• Failure to Capture Risk Reclassification: AUC does not measure whether individuals move across clinically relevant risk thresholds when new predictors are added to models, a critical consideration for stratified screening and prevention strategies [44].

• Dependence on Overall Model Performance: The interpretability of AUC changes (ΔAUC) diminishes as baseline model performance increases, making it difficult to evaluate PRS value when added to strong clinical predictors [44].

Table 1: Comparison of Metrics for Evaluating Polygenic Risk Model Improvements

Metric	What It Measures	Interpretation	Advantages	Limitations
ΔAUC	Improvement in discrimination between cases and controls	Higher values indicate better separation	Intuitive, widely understood	Insensitive to clinically important improvements
NRI (Net Reclassification Improvement)	Proportion of individuals reclassified into more appropriate risk categories	Positive values indicate improved reclassification	Captures movement across risk thresholds	Depends on predefined risk categories
IDI (Integrated Discrimination Improvement)	Improvement in average predicted risks between events and non-events	Positive values indicate better risk separation	Sensitive to risk magnitude changes	Less familiar to researchers; no universal benchmarks
Predictive R²	Proportion of variance explained by the model	Higher values indicate better fit	Direct interpretation; useful for power calculations	Depends on disease prevalence; not a discrimination measure

Practical Challenges in Endometriosis Research

Application of AUC to endometriosis subphenotype prediction faces specific methodological challenges:

• Heterogeneous Disease Manifestations: Endometriosis encompasses multiple subphenotypes (peritoneal, ovarian, deep infiltrating) with potentially distinct genetic architectures, complicating the interpretation of aggregate AUC values [46]. A model might demonstrate excellent discrimination for one subphenotype while performing poorly for others, yet report a deceptively adequate overall AUC.

• Sample Size Requirements: Detection of statistically significant AUC improvements requires large sample sizes, a particular challenge for rare endometriosis subphenotypes. For instance, one study predicting severe endometriosis achieved an AUC of 0.744 using a random forest model, but required 308 patients with surgical confirmation for development [47].

• Stage-Dependent Genetic Effects: Advanced-stage endometriosis (rASRM stage III/IV) demonstrates stronger genetic effects than earlier stages, suggesting that PRS performance may vary substantially across disease severity spectra [24]. AUC comparisons across studies that enroll different disease severity distributions can be misleading.

Complementary Metrics and Multidimensional Evaluation Frameworks

Beyond AUC: Integrated Discrimination and Reclassification Metrics

To address AUC limitations, researchers should incorporate complementary metrics that capture different dimensions of predictive performance:

• Net Reclassification Improvement (NRI): Quantifies the proportion of individuals correctly reclassified into higher or lower risk categories after adding PRS to a baseline model [44]. In practice, NRI calculation requires defining clinically meaningful risk thresholds specific to endometriosis subphenotypes (e.g., thresholds for surgical intervention, fertility preservation, or targeted medical therapy).

• Integrated Discrimination Improvement (IDI): Measures the average improvement in predicted risks between cases and controls, capturing increases in separation between event and non-event distributions [44]. IDI is particularly valuable for detecting improvements when risk distributions shift in ways not reflected in AUC changes.

• Calibration Metrics: Assessment of how well predicted probabilities match observed risks is crucial for clinical implementation. This can be evaluated using Hosmer-Lemeshow goodness-of-fit tests or calibration plots [48].

Table 2: Multimarker Assessment Strategies for Enhanced Endometriosis Prediction

Assessment Approach	Application in Endometriosis	Implementation Considerations
Multimarker Panels	Combining genetic, epigenetic, and protein biomarkers	Machine learning approaches (RF, XGBoost) effectively handle high-dimensional data [47] [1]
Clinical-Genetic Integration	Adding PRS to clinical risk factors (symptoms, imaging)	Requires standardized collection of clinical metadata [45] [46]
Disease Subphenotyping	Developing subtype-specific prediction models	Necessitates precise phenotypic characterization and sufficient sample sizes [24] [46]
Longitudinal Performance	Monitoring model performance across disease progression	Demands well-annotated longitudinal cohorts with repeated measures

Methodological Refinements for Enhanced Predictive Performance

Substantial improvements in predictive performance can be achieved through methodological innovations in both PRS construction and model development:

• Advanced PRS Methods: Novel approaches like EB-PRS that leverage effect size distributions across markers have demonstrated substantial improvements over standard PRS methods, with relative improvements in predictive R² ranging from 3.1% to 307.1% across various complex diseases [49]. These methods do not require external linkage disequilibrium reference panels or parameter tuning.

• Machine Learning Integration: Ensemble methods like random forest can capture non-linear relationships and complex interactions between genetic and clinical predictors. One study predicting severe pelvic endometriosis found that a random forest model incorporating both clinical and ultrasound features achieved the best performance (AUC = 0.744) among seven machine learning algorithms tested [47].

• Multi-ancestry PRS Development: Current PRS models predominantly reflect European-ancestry genetics, limiting generalizability. Developing multi-ancestry PRS (MA-PRS) that incorporate both disease-associated and ancestry-informative SNPs represents a critical direction for improving predictive power across diverse populations [45].

Experimental Protocols for Enhanced Predictive Modeling

Protocol 1: Comprehensive Model Evaluation Framework

Objective: To implement a multidimensional evaluation strategy for polygenic risk models of endometriosis subphenotypes that moves beyond sole reliance on AUC.

Materials:

• Genotyped cohort with confirmed endometriosis cases and controls
• Clinical and demographic metadata
• High-performance computing environment
• R or Python with specialized packages (PRSice2, pROC, nricens, PredictABEL)

Procedure:

• Base Model Development: Construct a clinical prediction model incorporating established risk factors (e.g., demographic characteristics, symptom profiles, imaging findings) [48].
• PRS Calculation: Generate polygenic risk scores using state-of-the-art methods (e.g., LDpred, EB-PRS, PRSice2) [49].
• Integrated Model Construction: Combine PRS with clinical predictors in a multivariable framework.
• Discrimination Assessment:
  • Calculate AUC for base, PRS-only, and integrated models
  • Compare AUC values using DeLong's test for paired comparisons
• Reclassification Analysis:
  • Define clinically relevant risk thresholds based on endometriosis management guidelines
  • Calculate categorical and continuous NRI
  • Compute IDI to assess improvement in risk separation
• Calibration Evaluation:
  • Generate calibration plots comparing predicted versus observed risks
  • Perform Hosmer-Lemeshow goodness-of-fit test
• Clinical Utility Assessment:
  • Conduct decision curve analysis to evaluate net benefit across risk thresholds
  • Stratify performance metrics by relevant subgroups (e.g., disease stage, age groups)

Expected Outcomes: This comprehensive protocol will determine whether PRS provides value beyond current prediction methods across multiple dimensions, not just discrimination.

Protocol 2: Advanced PRS Development with Effect Size Distribution Modeling

Objective: To implement advanced PRS methods that leverage effect size distributions for improved prediction of endometriosis subphenotypes.

Materials:

• GWAS summary statistics for endometriosis and relevant subphenotypes
• Individual-level genotype data for target cohort
• Reference panels matching ancestral background
• Computational resources for large-scale genetic analysis

Procedure:

• Data Preparation and QC:
  • Perform standard QC on GWAS summary statistics (filter for INFO score > 0.9, MAF > 0.01)
  • Apply stringent QC to target genotype data (call rate > 98%, HWE p > 1×10⁻⁶)
  • Ensure ancestral matching between discovery GWAS and target dataset
• Standard PRS Calculation:
  • Implement clumping and thresholding with multiple p-value thresholds (1, 0.5, 0.1, 0.05, 5×10⁻², 5×10⁻⁴, 5×10⁻⁶)
  • Select optimal p-value threshold via cross-validation
• EB-PRS Implementation:
  • Estimate empirical effect size distribution from GWAS summary statistics
  • Apply empirical Bayes shrinkage to effect sizes
  • Calculate posterior mean effect sizes for PRS construction [49]
• Comparative Evaluation:
  • Compare EB-PRS performance against standard PRS methods
  • Assess performance across endometriosis subphenotypes (superficial peritoneal, ovarian endometrioma, deep infiltrating)
  • Evaluate transferability across ancestral groups
• Functional Annotation Integration:
  • Incorporate tissue-specific functional annotations (e.g., endometrial epigenomic data)
  • Weight SNPs by biological priors using methods like AnnoPred

Expected Outcomes: Implementation of this protocol typically yields PRS with improved predictive accuracy compared to standard methods, particularly for complex subphenotypes with heterogeneous genetic architectures.

Table 3: Research Reagent Solutions for Enhanced Predictive Modeling

Category	Specific Resource	Function	Implementation Considerations
Genetic Data	GWAS summary statistics	Discovery of variant-trait associations	Ensure ancestral diversity; large sample sizes [45]
Genotyping Arrays	Illumina Global Screening Array	Genome-wide variant genotyping	Consider custom content for endometriosis-relevant loci
PRS Software	PRSice2, LDpred, LDPred2	Polygenic risk score calculation	LDpred requires LD reference panel [49]
Machine Learning Libraries	scikit-learn, mlr3, XGBoost	Advanced predictive modeling	Effective for clinical-genetic integration [47]
Bioinformatics Tools	PLINK, QCTOOL, R/Bioconductor	Genetic data processing and analysis	Standardized pipelines enhance reproducibility
Validation Cohorts	Deeply phenotyped endometriosis cohorts	Model validation and calibration	Must include relevant subphenotypes [24]

Moving beyond the limitations of AUC requires a fundamental shift in how we evaluate polygenic risk models for endometriosis subphenotypes. By implementing multidimensional assessment frameworks that incorporate reclassification metrics, calibration measures, and clinical utility analyses, researchers can more accurately characterize the value of genetic information for risk prediction. Methodological innovations in PRS construction, particularly approaches that leverage effect size distributions and incorporate functional genomic data, offer promising pathways for enhanced predictive power.

Future directions should prioritize multi-ancestry model development, integration of multi-omics data (epigenetic, transcriptomic, proteomic), and application of sophisticated machine learning approaches capable of capturing complex interactions. Furthermore, establishing standardized evaluation frameworks specific to endometriosis subphenotypes will enable more meaningful comparisons across studies and accelerate progress toward clinically implementable risk prediction tools. Through these coordinated advances, the field can overcome current limitations in predictive power and deliver on the promise of personalized risk assessment for this complex gynecological disorder.

Endometriosis is a multifaceted inflammatory disease with significant heterogeneity in its clinical presentation, encompassing three recognized lesion phenotypes (peritoneal, ovarian endometrioma, and deep infiltrating endometriosis) and diverse symptom profiles ranging from chronic pelvic pain to infertility [13] [50]. This clinical diversity presents substantial challenges for developing effective polygenic risk scores (PRS), as general PRS constructed for endometriosis as a single entity often fail to capture the genetic architecture underlying specific subphenotypes. The disease's complex pathophysiology involves interconnected mechanisms including hormonal dysregulation, immune dysfunction, oxidative stress, genetic and epigenetic alterations, and microbiome imbalances [13]. While PRS aggregates the effects of multiple genetic variants into a single risk measure, current evidence demonstrates that existing endometriosis PRS show limited utility in predicting specific clinical presentations, disease severity, or anatomical localization [29]. This application note examines the technical limitations of general PRS for endometriosis subphenotype prediction and provides detailed experimental protocols for developing more refined, phenotype-specific genetic risk tools.

Quantitative Assessment of Current Endometriosis PRS Limitations

Performance Evaluation of General PRS Across Studies

Table 1: Discriminatory Performance of Endometriosis PRS in Validation Studies

Cohort Description	Sample Size (Cases/Controls)	PRS Construction	OR per SD Increase	p-value	Subtype Analysis	Citation
Surgically confirmed cases (Western Danish Center)	249/348	14-SNP PRS	1.59	2.57×10-7	Ovarian: OR=1.72 (p=6.7×10-5); Infiltrating: OR=1.66 (p=2.7×10-9); Peritoneal: OR=1.51 (p=2.6×10-3)	[11]
Danish Twin Registry	140/316	14-SNP PRS	1.50	0.0001	Not reported	[11]
UK Biobank	2,967/256,222	14-SNP PRS	1.28	<2.2×10-16	Limited subtype differentiation	[11]
Swedish clinical cohort	172 (cases only)	13-SNP weighted PRS	Not significant	>0.05	Inverse association with spread (p-trend not significant)	[29]

Clinical Subphenotype Correlation Analysis

Table 2: Association Between PRS and Endometriosis Clinical Presentations

Clinical Characteristic	Association with PRS	Statistical Significance	Cohort	Implication	Citation
Spread of endometriosis	Inverse association	Lost significance when calculated as p for trend	Swedish cohort (N=172)	PRS not predictive of disease severity	[29]
Gastrointestinal tract involvement	Inverse association	Not significant	Swedish cohort (N=172)	Limited utility for predicting bowel endometriosis	[29]
Hormone treatment	Inverse association	Not significant	Swedish cohort (N=172)	Treatment response not genetically predicted	[29]
Ovarian endometriosis	Positive association	OR=1.72, p=6.7×10-5	Danish surgical cohort	Moderate predictive value for specific subtype	[11]
Infiltrating endometriosis	Positive association	OR=1.66, p=2.7×10-9	Danish surgical cohort	Better performance for infiltrating disease	[11]
Peritoneal endometriosis	Weakest association	OR=1.51, p=2.6×10-3	Danish surgical cohort	Limited utility for peritoneal disease	[11]

Molecular Complexity Underlying Endometriosis Subphenotypes

The limited performance of general PRS for specific presentations stems from the diverse molecular mechanisms driving different endometriosis subphenotypes. Recent multi-omics analyses reveal distinct pathways contribute to the disease heterogeneity:

Hormonal Dysregulation Pathways

Endometriosis exhibits local estrogen dominance despite normal circulating levels, driven by overexpression of aromatase (CYP19A1) and downregulation of 17β-hydroxysteroid dehydrogenase type 2 in ectopic lesions. Concurrent progesterone resistance results from reduced PR-B isoform expression due to promoter hypermethylation and microRNA dysregulation (e.g., miR-26a, miR-181) [13]. These hormonal variations differ across subphenotypes, contributing to PRS inaccuracies.

Immune and Inflammatory Mechanisms

Pervasive immune dysregulation characterizes endometriosis, with macrophages constituting over 50% of immune cells in peritoneal fluid and exhibiting impaired phagocytic activity due to downregulated CD36 expression. Alterations in natural killer (NK) cell cytotoxicity and T-cell subset dysregulation (increased Th2, Th17, and Treg cells) vary across disease presentations [13]. This immunological heterogeneity is not captured by general PRS.

Genetic and Epigenetic Factors

Beyond common variants included in PRS, epigenetic modifications such as N6-methyladenosine (m6A) methylation regulators (HNRNPA2B1 and HNRNPC) serve as potential biomarkers for endometriosis-related infertility [51]. These epigenetic mechanisms contribute to subphenotype specificity but are not incorporated into current PRS models.

Experimental Protocols for Advanced PRS Development

Protocol 1: Pathway-Specific Polygenic Risk Score (pPRS) Construction

Background: Standard PRS includes variants that affect disease risk independent of environmental factors, diluting signals from variants involved in specific pathways. Pathway PRS (pPRS) focuses on biologically relevant variant subsets to improve subphenotype prediction [52].

Materials:

• GWAS summary statistics for endometriosis (minimum 15,000 cases)
• Genotype data from target cohort (minimum 2,000 cases with subphenotype data)
• Functional annotation databases (ENCODE, Roadmap Epigenomics, Genotype-Tissue Expression)
• Pathway analysis tools (GARFIELD, DEPICT, Pascal)

Procedure:

• Variant Annotation: Annotate GWAS variants using functional genomic data from relevant tissues (endometrium, ovarian tissue, immune cells)
• Pathway Mapping: Assign variants to biological pathways using curated gene sets (KEGG, Reactome, GO)
• pPRS Calculation:
  • Compute separate PRS for each pathway using PRSice-2 or LDPred
  • Apply Bayesian methods for effect size shrinkage (SBayesR)
• Interaction Testing:
  • Test pPRS × subphenotype interactions using multivariate regression
  • Validate significant interactions in independent cohort

Validation: Assess pPRS discrimination for specific subphenotypes using ROC analysis and calculate net reclassification improvement compared to general PRS.

Protocol 2: Integration of PRS with Multi-Omics Data for Subphenotype Prediction

Background: Integrating genetic risk with transcriptomic, epigenomic, and proteomic data can capture the molecular diversity of endometriosis subphenotypes [13] [51].

Materials:

• Blood and endometriosis tissue samples (minimum 100 per subphenotype)
• RNA sequencing equipment (Illumina NovaSeq)
• Methylation array platform (Infinium MethylationEPIC)
• Multiplex immunoassay system (Olink Inflammation panel)
• Genotyping array (Global Screening Array)

Procedure:

• Multi-Omics Data Generation:
  • Perform RNA-seq on ectopic and eutopic endometrial tissues
  • Conduct DNA methylation profiling using bisulfite sequencing
  • Quantify inflammatory proteins (OSM, MCP-1, TNFRSF9) using Proseek Multiplex assay
• Data Integration:
  • Identify subtype-specific molecular signatures using differential expression analysis
  • Construct multi-omics similarity networks for subphenotype classification
  • Build integrative models using regularized regression (elastic net)
• PRS Enhancement:
  • Weight PRS variants by functional genomic annotations
  • Develop subphenotype-specific PRS using omics-informed priors

Validation: Perform cross-validation within cohort and external validation in independent population. Compare subphenotype classification accuracy against clinical assessment.

Protocol 3: Neural Network-Based PRS with Functional Annotations

Background: Traditional PRS methods assume linear additive effects, potentially missing non-linear relationships between genotypes and subphenotypes. Neural network approaches can learn complex annotation-function relationships [53].

Materials:

• Curated functional annotations (chromatin accessibility, TF binding, sequence conservation)
• High-performance computing cluster with GPU acceleration
• UK Biobank or comparable genetic dataset with endometriosis cases
• Python with TensorFlow/PyTorch and PRS analysis libraries

Procedure:

• Annotation Curation:
  • Compile extensive functional annotations including ancestry-stratified allele frequencies
  • Include chromatin accessibility across reproductive tissue cell types
  • Incorporate transcription factor binding from ENCODE and quantitative trait loci
• Model Architecture:
  • Implement neural network with SNP annotations as input
  • Use empirical Bayesian framework to learn annotation-weight relationships
  • Incorporate non-linear activation functions to capture complex relationships
• Model Training:
  • Train on endometriosis GWAS summary statistics
  • Optimize hyperparameters using Bayesian optimization
  • Regularize to prevent overfitting (dropout, weight decay)
• Subphenotype Application:
  • Calculate PRSFNN scores for each subphenotype
  • Assess stratification performance across clinical presentations

Validation: Benchmark against other PRS methods (LDpred, PRS-CS) using out-of-sample validation. Evaluate improvement in subphenotype classification accuracy.

The Scientist's Toolkit: Essential Research Reagents and Platforms

Table 3: Key Research Reagents and Platforms for Endometriosis Subphenotype PRS Development

Category	Specific Tool/Reagent	Application in PRS Research	Key Features	Representative Use
Genotyping Platforms	Illumina Global Screening Array	Genome-wide variant detection	~650,000 markers optimized for imputation	Initial genotyping in PRS studies [29]
	Color Health NGS panel	Targeted sequencing for PRS	Customizable SNP content (75-126 SNPs)	PRS implementation in WISDOM trial [54]
Functional Genomics	Proseek Multiplex Inflammation panel (Olink)	Inflammation protein quantification	92 inflammatory proteins, high sensitivity	Mapping molecular subphenotypes [29]
	ENCODE4 database	Transcription factor binding annotation	1,600+ experiments across cell types	Functional annotation for neural network PRS [53]
Bioinformatics Tools	PRSice-2	PRS calculation and validation	Fast, efficient, clumping and thresholding	Standard PRS development [11]
	SBayesR (GCTB)	Bayesian PRS optimization	Sparse effects modeling, improves prediction	Enhanced effect size estimation [9]
	FlashPCA	Population stratification control	Efficient principal component analysis	Ancestry adjustment in PRS [29]
Validation Assays	Visual Analog Scale for IBS (VAS-IBS)	Bowel symptom quantification	Patient-reported outcome measure	GI subphenotype characterization [29]
	ENDOGRAM tissue classification	Molecular subtyping	Histopathological and molecular analysis	Linking pathology to genetics [50]

The development of subphenotype-specific PRS for endometriosis requires a paradigm shift from general genetic risk assessment to integrated multi-omics approaches. Current evidence demonstrates that general endometriosis PRS, while informative for overall disease risk, lack precision for predicting specific clinical presentations, anatomical localizations, or treatment responses. The future of endometriosis PRS development lies in pathway-specific approaches, neural network methods incorporating functional annotations, and sophisticated integration of genetic data with transcriptomic, epigenomic, and proteomic profiles. These advanced methodologies promise to bridge the current subphenotype prediction gaps, ultimately enabling personalized risk prediction and targeted interventions for this heterogeneous disease. Implementation of the detailed experimental protocols outlined in this application note will accelerate progress toward clinically useful subphenotype prediction tools for endometriosis management.

The development of polygenic risk scores (PRS) for endometriosis represents a significant advance in understanding the genetic architecture of this complex condition. However, a substantial limitation impedes their broader application: current PRS models are predominantly derived from genome-wide association studies (GWAS) conducted in populations of European ancestry (EUR) [55] [56]. This creates a critical global imbalance in precision medicine, as PRS generated from GWAS in one population typically provide attenuated predictive accuracy when applied to other populations [57] [56]. The transferability challenge arises from multiple factors, including differences in linkage disequilibrium (LD) patterns between populations, allele frequency differences, SNP array design biases toward European variants, and the limited representation of diverse populations in genetic research cohorts [57] [55]. Until recently, over 80% of participants in genetic studies were of European descent, with only approximately 4% of GWAS participants representing East Asian ancestry [57] [55]. This bias risks exacerbating health disparities if clinically implemented, as PRS may misestimate genetic risk for individuals of non-European ancestry [57]. This Application Note provides a comprehensive framework for addressing population-specific biases in endometriosis PRS development, enabling more equitable precision medicine approaches across diverse populations.

Quantitative Assessment of Current Limitations

Table 1: PRS Performance Comparison Across Ancestral Groups

Ancestral Group	GWAS Representation	PRS Transfer Performance	Key Limiting Factors
European	~80% of participants [57]	Reference standard (OR = 1.28-1.59 for endometriosis) [11]	Baseline reference
East Asian	~4% of participants [55]	Moderately reduced	LD differences, allele frequency spectra
African	<3% of participants [56]	Severely reduced	Greater genetic diversity, limited reference panels
Admixed Populations	Highly underrepresented	Unpredictable biases [57]	Differential genetic drift, complex ancestry

Table 2: Endometriosis-Specific Genetic Discovery by Ancestry

Parameter	European Ancestry	East Asian Ancestry	African Ancestry
Sample Size in Largest GWAS	~60,674 cases [15]	Limited representation in international consortia	Minimal representation
Number of Identified Loci	42 genome-wide significant loci [15]	Data insufficient for comparison	Data unavailable
Variance Explained	Up to 5.01% [15]	Expected reduction in transferred PRS	Expected significant reduction
Population-Specific Variants	14-SNP PRS developed [11]	Potential undiscovered variants	Likely numerous undiscovered variants

Protocols for Developing Population-Aware PRS

Protocol: Multi-Ancestry GWAS Implementation for Endometriosis

Purpose: To identify population-specific and shared genetic risk variants for endometriosis across diverse ancestral groups.

Materials:

• Biological samples from globally diverse populations
• Custom genotyping arrays optimized for diverse populations or whole-genome sequencing
• High-performance computing infrastructure
• Population reference panels (1000 Genomes, gnomAD, population-specific references)

Procedure:

• Cohort Selection: Recruit participants from multiple ancestral backgrounds using standardized ancestry informative markers [55].
• Genotyping Platform Selection: Utilize cosmopolitan arrays or low-coverage sequencing (<1× depth) to reduce ascertainment bias [57].
• Quality Control: Apply ancestry-specific filters for Hardy-Weinberg equilibrium, missingness, and minor allele frequency.
• Association Testing: Perform GWAS stratified by genetic ancestry using linear mixed models to account for population structure.
• Meta-Analysis: Conduct trans-ancestry meta-analysis using fixed-effects or random-effects models based on heterogeneity statistics [8].
• Variant Annotation: Functionally characterize associated variants using eQTL data from diverse tissues (uterus, ovary, blood) [7].

Troubleshooting:

• For heterogeneous genetic effects across populations, consider subgroup analyses by disease stage (rASRM III/IV) as effect sizes are consistently greater in advanced disease [8] [24].
• When sample sizes are unequal across groups, apply statistical methods to prevent larger cohorts from dominating association signals.

Protocol: Population-Specific PRS Construction and Calibration

Purpose: To develop and optimize PRS for endometriosis in underrepresented populations.

Materials:

• Summary statistics from multi-ancestry GWAS
• Target cohort with genetic and phenotypic data
• LD reference panels matched to target population
• PRS computation software (PRSice, PLINK, LDpred, SBayesR)

Procedure:

• Base Data Preparation: Process GWAS summary statistics using population-specific LD reference panels [9].
• PRS Method Selection: Apply Bayesian methods (SBayesR) for improved cross-population prediction [9].
• Score Calculation: Generate PRS using clumping and thresholding or LD-based methods.
• Calibration: Adjust effect sizes using ancestry-specific scaling factors [55].
• Validation: Assess PRS performance in independent validation cohorts of matched ancestry.
• Clinical Translation: Convert PRS percentiles to absolute disease risk using ancestry-specific prevalence estimates [56].

Troubleshooting:

• For limited sample sizes in target populations, consider cross-population polygenic prediction methods that leverage information across ancestries [56].
• When transferring European-derived PRS to other populations, account for ancestral vs. derived allele status to reduce bias [57].

Signaling Pathways and Workflow Visualization

Diagram 1: Comprehensive workflow for developing ancestry-aware polygenic risk scores for endometriosis, highlighting parallel analysis pathways across diverse populations.

Diagram 2: Key sources of population-specific bias in PRS development and corresponding mitigation strategies.

The Scientist's Toolkit

Table 3: Essential Research Reagents and Computational Tools

Tool/Reagent	Specification	Application in Endometriosis PRS
Cosmopolitan SNP Arrays	Designed with variants informative across diverse populations	Reduces genotyping ascertainment bias in multi-ancestry cohorts [57]
GTEx Database v8	Tissue-specific eQTL data from uterus, ovary, blood [7]	Functional characterization of endometriosis risk variants across ancestries
LD Reference Panels	Population-specific (AFR, EAS, EUR) from 1000 Genomes	Improves PRS accuracy in target populations [56]
GCTB Software	Implements SBayesR method	Bayesian approach for PRS construction with improved cross-population performance [9]
PRSice-2	Clumping and thresholding algorithm	Computes and validates PRS across multiple p-value thresholds [56]
METAL Software	Trans-ancestry meta-analysis	Combines GWAS results across diverse cohorts with heterogeneity testing [8]

Discussion and Future Directions

The Taiwan Precision Medicine Initiative (TPMI) demonstrates the substantial benefits of population-specific genomic research, having developed PRS for Han Chinese ancestry that account for up to 10.3% of health variation in that cohort [55]. This initiative identified 95 new genetic associations that were previously undetected in European-focused studies, primarily due to allele frequency differences and population-specific genetic effects [55]. Similarly, for endometriosis research, expanding GWAS in diverse populations may reveal ancestry-specific variants in genes like WNT4, VEZT, and GREB1, which have established roles in endometriosis pathogenesis [8] [1].

Future directions should prioritize the development of "polyethnic" scores that optimally combine trans-ethnic and ethnic-specific information [56]. Methods like XP-BLUP and multi-ethnic PRS are showing promise in improving predictive accuracy across diverse populations [56]. Additionally, integrating functional genomics data, including endometrial DNA methylation quantitative trait loci (mQTLs) and tissue-specific regulatory elements, will enhance the biological interpretation of population-specific risk variants [24]. As these approaches mature, researchers must simultaneously address non-genetic sources of health disparities, including healthcare access, environmental exposures, and social determinants of health, to ensure equitable benefits from precision medicine advances in endometriosis care [57].

Endometriosis is a complex, chronic inflammatory gynecological disease affecting approximately 10% of women of reproductive age worldwide and is found in 30-50% of women undergoing infertility evaluation [58] [13]. Its pathogenesis involves a multifactorial etiology with an estimated 50% heritable genetic component and 50% contribution from environmental factors, which often manifest through epigenetic modifications such as DNA methylation [59]. The disease demonstrates remarkable heterogeneity in clinical presentation, lesion distribution, and molecular profiles, complicating both diagnosis and treatment [14].

The integration of multi-omics data represents a transformative approach for deciphering this complexity. By combining methylation risk scores (MRS) with protein biomarker signatures, researchers can now develop more precise classification systems that correlate molecular subphenotypes with clinical manifestations and therapeutic responses [60] [14]. This application note provides detailed protocols for generating and validating these multi-omics signatures within the context of polygenic risk score development for endometriosis subphenotypes.

Background and Significance

Endometriosis Pathophysiology and Molecular Heterogeneity

Endometriosis pathophysiology involves several interconnected mechanisms that create a hostile reproductive environment:

• Hormonal Dysregulation: Local estrogen dominance with progesterone resistance characterizes endometriotic lesions, facilitated by epigenetic alterations in estrogen receptor (ERβ) and aromatase promoters [58] [13]
• Immune Dysfunction: Aberrant immune cell activation, including macrophage polarization shifts and impaired natural killer cell cytotoxicity, promotes chronic inflammation [58] [13]
• Oxidative Stress and Ferroptosis: Iron-driven oxidative stress particularly injures granulosa cells, impacting oocyte competence [58] [13]
• Microbiome Imbalance: Reproductive tract and gut microbiome dysbiosis modulates local estrogen metabolism and inflammation [58] [13]

The molecular heterogeneity of endometriosis is evident across different lesion types—superficial peritoneal endometriosis, ovarian endometriomas, and deep infiltrating endometriosis—each demonstrating distinct transcriptional and epigenetic profiles [14].

Current Diagnostic Challenges and the Need for Biomarkers

Current endometriosis diagnosis relies on surgical visualization with histologic confirmation, resulting in an average diagnostic delay of 7-11 years from symptom onset [14] [59]. This diagnostic lag allows disease progression and potentially irreversible damage to reproductive organs. The development of non-invasive biomarkers using multi-omics approaches represents an urgent unmet clinical need that could enable earlier intervention and personalized treatment strategies [14] [61].

Multi-Omics Integration Framework

Conceptual Workflow for Multi-Omics Data Integration

The following diagram illustrates the comprehensive workflow for integrating multi-omics data to develop refined endometriosis subphenotypes:

This integrated approach enables researchers to move beyond traditional classification systems (rASRM, ENZIAN) toward molecularly-defined subphenotypes with distinct clinical trajectories and therapeutic responses [60] [14].

Key Signaling Pathways in Endometriosis

The pathophysiology of endometriosis involves dysregulation of several key signaling pathways, many influenced by epigenetic modifications:

DNA methylation modifications in endometriosis affect genes involved in these critical pathways, contributing to disease establishment and progression [59]. Notably, hypomethylation of ESR2 (encoding ERβ) and aromatase promoters enhances local estrogen production, while hypermethylation of progesterone receptor promoters drives progesterone resistance [58] [13] [59].

Methylation Risk Score Modeling

MRS Development Protocol

Objective: Develop a methylation risk score (MRS) for endometriosis classification using endometrial tissue methylation data.

Sample Requirements:

• Cases: 590 endometriosis patients with surgically/histologically confirmed diagnosis
• Controls: 318 women without endometriosis
• Tissue: Endometrial biopsy collected during proliferative or secretory phase
• Storage: Snap-frozen in liquid nitrogen within 30 minutes of collection [60]

Experimental Workflow:

Detailed Methodology:

• DNA Extraction and Bisulfite Conversion

  • Extract genomic DNA using QIAamp DNA Mini Kit (Qiagen)
  • Treat 500ng DNA with EZ-96 DNA Methylation-Lightning MagPrep (Zymo Research) following manufacturer's protocol
  • Assess conversion efficiency with internal controls [60]

• Methylation Profiling

  • Utilize Illumina EPIC methylation arrays covering >850,000 CpG sites
  • Process arrays according to standard Illumina protocols
  • Include technical replicates and control samples in each batch [60]

• Quality Control and Normalization

  • Remove probes with detection p-value >0.01 in >10% samples
  • Exclude cross-reactive probes and those containing SNPs
  • Normalize data using functional normalization (FunNorm) or quantile normalization
  • Correct for batch effects using ComBat or SVA [60]

• MRS Model Construction

  • Apply elastic net regression with 10-fold cross-validation on training set (70% samples)
  • Include age, institution, and genetic ancestry as covariates
  • Select optimal lambda parameter minimizing cross-validation error
  • Calculate MRS as weighted sum of significantly associated CpG sites [60]

Performance Metrics:

• Best-performing MRS achieved AUC of 0.6748 using 746 CpG sites
• Combined MRS and PRS outperformed PRS alone in classification accuracy
• MRS captured 12-19.58% of variance in endometriosis status independent of genetic effects [60]

MRS Analytical Validation

Table 1: Methylation Risk Score Performance Characteristics

Parameter	Value	Description
Sample Size	908 individuals	590 cases, 318 controls
Optimal CpG Panel	746 sites	Selected via elastic net regression
Area Under Curve (AUC)	0.6748	Classification performance
Variance Captured	12-19.58%	Independent of common genetic variants
Covariates	Age, institution, genetic ancestry	Included in final model
Validation Approach	Train-test split by institution	Prevents overfitting

The MRS demonstrates that DNA methylation profiles in endometrial tissue provide significant predictive value for endometriosis classification beyond genetic factors alone [60]. This epigenetic component likely reflects the environmental contributions to endometriosis risk and progression.

Protein Biomarker Discovery and Validation

Proteomic Signature Workflow

Objective: Identify and validate protein biomarkers in serum/plasma that complement MRS for endometriosis subphenotyping.

Sample Requirements:

• Cases: 38 endometriosis patients (ASRM stages I-IV)
• Controls: 40 healthy women without pelvic pain or inflammation
• Sample Type: Serum collected in SST tubes, processed within 2 hours
• Storage: Aliquot and store at -80°C until analysis [61]

Experimental Workflow:

Detailed Methodology:

• Sample Preparation

  • Deplete high-abundance proteins (albumin, IgG) using ProteoPrep Immunoaffinity Albumin and IgG Depletion Kit (Sigma-Aldrich)
  • Reduce with 5mM DTT (30min, 60°C), alkylate with 15mM iodoacetamide (30min, RT in dark)
  • Digest with sequencing-grade trypsin (1:50 enzyme:protein, 16h, 37°C)
  • Desalt using C18 solid-phase extraction cartridges [61]

• LC-MS/MS Analysis

  • Separate peptides using Dionex Ultimate 3000 RSLC nanoSystem with PepMap C18 column (75μm × 50cm, 2μm)
  • Perform 120min gradient from 2-30% acetonitrile in 0.1% formic acid
  • Analyze with Orbitrap Fusion Lumos Mass Spectrometer
  • Acquire data in data-dependent acquisition mode with 3sec cycle time [61]

• Data Processing

  • Search raw files against human UniProt database using MaxQuant (v2.0.3)
  • Set false discovery rate (FDR) to 1% at protein and peptide level
  • Normalize protein intensities using MaxLFQ algorithm
  • Perform statistical analysis with Perseus or LIMMA [61]

• Biomarker Validation

  • Select top candidate proteins (5-10) for orthogonal validation
  • Develop ELISA assays using matched serum samples
  • Assess diagnostic performance using ROC curve analysis [61]

Cell-Free DNA and Methylation Biomarkers

Cell-Free DNA Quantification Protocol:

• Extract cf-DNA from 1mL serum using QIAamp Circulating Nucleic Acid Kit (Qiagen)
• Quantify using fluorometric methods (Qubit dsDNA HS Assay)
• Analyze differential methylation of 9 candidate gene promoters via bisulfite sequencing
• Calculate composite score combining cf-DNA concentration and methylation signature [61]

Key Findings:

• Endometriosis patients showed 3.9-fold higher cf-DNA levels versus controls
• Nine-gene methylation signature provided additional discriminatory power
• Combined approach achieved sensitivity of 70%, specificity of 87% for minimal-mild disease [61]

Table 2: Protein and cf-DNA Biomarker Performance

Biomarker Type	Analytical Platform	Key Findings	Performance
Cell-Free DNA	QIAamp Circulating Nucleic Acid Kit	3.9x higher in endometriosis vs controls	Sensitivity 70%, Specificity 87%
Methylation Signature	Targeted bisulfite sequencing	9 genes with differential methylation	Improved classification when combined with cf-DNA
Proteomic Profile	LC-MS/MS	Multiple inflammatory markers elevated	Complementary to epigenetic markers

Integrated Data Analysis and Subphenotyping

Multi-Omics Integration Protocol

Objective: Integrate MRS, PRS, and protein biomarkers to define molecular subphenotypes of endometriosis.

Computational Workflow:

• Data Preprocessing

  • Standardize all biomarkers to z-scores
  • Impute missing values using k-nearest neighbors (k=10)
  • Apply ComBat for batch correction across datasets

• Dimension Reduction

  • Perform multi-block partial least squares discriminant analysis (MB-PLS-DA)
  • Apply multi-omics factor analysis (MOFA+) to identify latent factors
  • Visualize using uniform manifold approximation and projection (UMAP)

• Subphenotype Identification

  • Apply consensus clustering across omics layers
  • Use k-means clustering (k=3-5) on integrated factor matrix
  • Validate clusters using silhouette width and stability measures

• Clinical Correlation

  • Associate subphenotypes with clinical variables (pain symptoms, infertility, disease stage)
  • Evaluate treatment response differences across subphenotypes
  • Assess time to recurrence by cluster assignment

Validation Approach:

• Split data into discovery (70%) and validation (30%) sets
• Apply bootstrap resampling to assess cluster stability
• Validate in independent cohort when available

Research Reagent Solutions

Table 3: Essential Research Reagents for Multi-Omics Endometriosis Studies

Category	Specific Product	Application	Key Features
DNA Methylation	Illumina Infinium MethylationEPIC Kit	Genome-wide methylation profiling	>850,000 CpG sites, comprehensive coverage
Bisulfite Conversion	Zymo Research EZ-96 DNA Methylation-Lightning MagPrep	DNA treatment for methylation analysis	Rapid 90min protocol, >99% conversion efficiency
cf-DNA Extraction	QIAamp Circulating Nucleic Acid Kit (Qiagen)	Isolation from serum/plasma	Optimized for low-abundance circulating DNA
Protein Depletion	ProteoPrep Immunoaffinity Albumin and IgG Depletion Kit	Serum proteome simplification	Removes >95% abundant proteins
Mass Spectrometry	Orbitrap Fusion Lumos Tribrid Mass Spectrometer	Proteomic quantification	High sensitivity and resolution
Data Analysis	MOFA+ (Multi-Omics Factor Analysis)	Multi-omics integration	Identifies latent factors across data types

Application in Drug Development

The integration of MRS and protein biomarkers offers significant promise for enriching clinical trials and personalizing therapeutic approaches:

• Patient Stratification: MRS can identify patients with specific epigenetic profiles who may respond better to targeted therapies [62]
• Treatment Response Prediction: Protein biomarkers can monitor molecular responses to interventions before clinical improvements manifest [62]
• Clinical Trial Enrichment: Multi-omics signatures can identify patients most likely to benefit from investigational therapies, increasing trial efficiency [62]

Recent analyses of FDA submissions demonstrate increasing use of polygenic risk scores in early-phase clinical trials, particularly in neurology, oncology, and psychiatry [62]. This trend is now extending to endometriosis with the development of validated MRS and protein signatures.

The integration of methylation risk scores and protein biomarkers represents a powerful approach for deciphering endometriosis heterogeneity and advancing personalized medicine. The protocols detailed in this application note provide a roadmap for generating validated multi-omics signatures that can classify disease subphenotypes, predict treatment responses, and identify novel therapeutic targets.

As these technologies mature, multi-omics profiling is poised to transform endometriosis management—reducing diagnostic delays, enabling targeted interventions, and ultimately improving reproductive outcomes for the millions of women affected by this complex condition.

Endometriosis demonstrates profound clinical heterogeneity, with presentation varying from superficial peritoneal lesions to deeply infiltrating disease and ovarian endometriomas [63] [46]. This heterogeneity represents a significant challenge in genetic studies, as traditional genome-wide association studies (GWAS) that treat endometriosis as a single entity have explained only approximately 2.2% of disease variance despite an estimated heritability of ~50% [63] [64]. The limited observed heritability in large genetic association studies may be attributable to underlying heterogeneity of disease mechanisms, creating critical statistical power constraints that necessitate sophisticated approaches to subphenotype characterization and sample size determination [65] [64].

Emerging evidence suggests that different endometriosis subtypes likely have distinct genetic architectures. Interim results from a large meta-analysis identified 27 genome-wide significant loci, with 78% demonstrating greater effect sizes in stage III/IV disease compared to stage I/II, and 63% showing greater effect sizes in endometriosis with infertility [63]. This genetic heterogeneity underscores the necessity of well-powered subphenotype studies to uncover the full spectrum of endometriosis risk variants and facilitate meaningful polygenic risk score development.

Quantitative Framework: Sample Size Requirements for Subphenotype Studies

Established Benchmarks in Endometriosis Genetic Research

Table 1: Sample Size Benchmarks in Recent Endometriosis Genetic Studies

Study Focus	Total Sample Size	Cases	Controls	Key Findings	Reference
Multi-ancestry GWAS	~1.4 million participants	105,869	~1.3 million	80 genome-wide significant associations (37 novel)	[27]
Indian population study	4,000	2,000	2,000	Aimed to address representation gap in global consortia	[63]
Clinical subphenotype clustering	12,350 cases	12,350	466,261	5 distinct subphenotype clusters with specific genetic associations	[65] [64]
DNA methylation analysis	984	637	347	15.4% of endometriosis variation captured by DNA methylation	[24]

Power Considerations for Subphenotype Analyses

The sample size requirements for endometriosis subphenotype studies are substantially influenced by several key factors:

• Minor Allele Frequency: Rare variants (MAF < 0.01) require dramatically larger sample sizes for adequate power compared to common variants
• Effect Size: Odds ratios below 1.2 necessitate sample sizes in the tens of thousands for adequate detection
• Genetic Architecture: Differences in effect sizes across subphenotypes impact power calculations
• Population Diversity: Transferability of polygenic risk scores across ancestries requires inclusion of diverse populations [27]

Recent research indicates that for well-powered subphenotype analyses, individual clusters should ideally contain at least 1,000 cases to detect moderate genetic effects (OR > 1.3) for common variants [65] [64]. The identification of five clinical subphenotype clusters in electronic health record data demonstrates how subphenotype stratification can enhance genetic discovery, with each cluster showing distinct genetic associations including PDLIM5 for pain comorbidities, GREB1 for uterine disorders, and WNT4 for pregnancy complications [64].

Methodological Approaches for Subphenotype Characterization

Standardized Phenotyping Protocols

Table 2: Endometriosis Subphenotype Classification Systems

Classification System	Subphenotypes Identified	Application in Genetic Studies	Strengths
rASRM Surgical Staging	Stages I-IV based on lesion appearance and extent	GWAS stratification by disease severity	Widely adopted, standardized scoring
Lesion Type Classification	Superficial Peritoneal (SUP), Ovarian Endometrioma (OMA), Deep Infiltrating Endometriosis (DIE)	Differential genetic effect sizes across subtypes	Direct mapping to pathological processes
Clinical Symptom Clustering	Pain comorbidities, Uterine disorders, Pregnancy complications, Cardiometabolic comorbidities, Asymptomatic	EHR-based clustering for genetic association	Captures clinical heterogeneity beyond surgical findings
Genital vs. Extragenital	Reproductive organ involvement vs. non-reproductive organ involvement	Understanding somatic mutation patterns	Accounts for lesion location heterogeneity

Unsupervised Clustering Protocol for Subphenotype Identification

Objective: To identify clinically meaningful endometriosis subphenotypes from electronic health record data for enhanced genetic association power.

Materials:

• Electronic Health Record system with longitudinal data
• ICD code mappings for endometriosis (N80 in ICD-10) and related conditions
• Computational infrastructure for high-dimensional data clustering

Procedure:

• Cohort Identification: Extract all patients with endometriosis diagnosis codes (ICD-10 N80)
• Feature Selection: Calculate prevalence of clinical features including:
  • Pain symptoms (dysmenorrhea, dyspareunia, chronic pelvic pain)
  • Comorbid conditions (migraine, IBS, fibromyalgia, asthma)
  • Fertility indicators (infertility diagnoses, pregnancy complications)
  • Surgical findings and lesion locations
• Dimensionality Reduction: Apply principal component analysis to clinical feature matrix
• Cluster Optimization: Test multiple clustering algorithms (k-means, spectral clustering, hierarchical clustering) and cluster numbers (K=2-20) using distortion curves and cluster separation metrics
• Cluster Validation: Characterize clusters through prevalence tests of clinical features compared to other clusters
• Genetic Association: Perform association testing of known endometriosis loci within each cluster

Expected Outcomes: Identification of 5 distinct subphenotype clusters with characteristic clinical profiles and differential genetic associations [64].

Visualization: Subphenotype Study Design Workflow

Subphenotype Study Design Workflow: This diagram illustrates the comprehensive approach from patient recruitment through genetic discovery, emphasizing the iterative process of phenotypic characterization and genetic validation necessary for well-powered subphenotype studies.

Advanced Integration of Multi-Omic Data

Methylation Quantitative Trait Loci (mQTL) Mapping Protocol

Objective: To identify epigenetic regulation of endometriosis risk through methylation quantitative trait loci analysis.

Materials:

• Endometrial tissue samples from cases and controls (n > 900 recommended)
• Illumina Infinium MethylationEPIC BeadChip kits
• Genotyping arrays or whole-genome sequencing data
• Bioinformatics pipelines for mQTL analysis

Procedure:

• Sample Collection: Obtain eutopic endometrial samples with detailed menstrual cycle phase documentation
• DNA Extraction: Isolve genomic DNA using standardized protocols
• Methylation Array Processing: Process samples through Illumina MethylationEPIC BeadChip following manufacturer protocols
• Quality Control: Remove probes with detection p-value > 0.01, beadcount < 3, or cross-reactive probes
• Normalization: Perform functional normalization using control probes
• Covariate Adjustment: Account for technical variables (batch, array position) and biological covariates (age, menstrual cycle phase)
• mQTL Mapping: Test associations between genetic variants and methylation levels within 1Mb of CpG sites
• Integration with GWAS: Colocalization analysis to identify shared causal variants between mQTLs and endometriosis risk loci

Expected Outcomes: Identification of approximately 118,185 independent cis-mQTLs including 51 associated with endometriosis risk, as demonstrated in recent large-scale analyses [24].

Research Reagent Solutions for Endometriosis Subphenotyping

Table 3: Essential Research Reagents for Endometriosis Subphenotype Studies

Reagent/Category	Specific Examples	Application	Function in Research
DNA Collection	EDTA blood collection tubes, DNA extraction kits (Qiagen, Thermo Fisher)	Genetic variant identification	High-quality DNA for genotyping and sequencing
Methylation Analysis	Illumina Infinium MethylationEPIC BeadChip	Epigenetic profiling	Genome-wide DNA methylation quantification at >850,000 sites
Protein Biomarkers	ELISA kits (Human R-Spondin3 ELISA Kit)	Protein quantification	Validation of proteomic findings from pQTL studies
Tissue Processing	TRIzol reagent, RNAlater	Transcriptomic analysis	RNA preservation and extraction for gene expression studies
Single-Cell Technologies	10x Genomics Chromium System, dissociation enzymes	Cellular heterogeneity characterization	Resolution of cell-type specific signatures in lesions
Immunoassays	Multiplexed immunoaffinity assays (SOMAscan)	Plasma protein measurement	High-throughput proteomic profiling for pQTL studies

Statistical Power Calculation Framework

Power Analysis Considerations for Subphenotype Studies

Adequate statistical power for endometriosis subphenotype studies requires careful consideration of multiple factors:

• Case-Control Ratio: Optimal power is typically achieved with 1:4 to 1:10 case-control ratios for common variants
• Multiple Testing Correction: Bonferroni correction for 5 subphenotypes increases significance threshold to p < 1×10⁻⁹
• Genetic Effect Heterogeneity: Sample size must account for potentially different genetic effect sizes across subphenotypes
• Ancestry Diversity: Limited transferability of European-derived PRS to other ancestries necessitates ancestry-specific sampling [27]

For rare subphenotypes (prevalence < 5% in endometriosis population), sample sizes exceeding 50,000 total cases may be required to detect moderate genetic effects (OR > 1.5). The recent identification of five endometriosis subphenotype clusters through EHR data mining demonstrated that cluster-specific genetic associations could be detected with cluster sizes ranging from 441 to 1,151 in a discovery cohort of 4,078 cases [64].

Addressing statistical power constraints in endometriosis subphenotype studies requires multi-faceted strategies including international collaborations to achieve sufficient sample sizes, standardized phenotyping protocols to reduce heterogeneity, innovative clustering approaches to identify biologically meaningful subgroups, and integration of multi-omic data to enhance discovery power. The development of robust polygenic risk scores for endometriosis subphenotypes depends on overcoming these power constraints through carefully designed studies that acknowledge and account for the substantial clinical and genetic heterogeneity of this complex disease.

Future directions should prioritize diverse population inclusion, longitudinal phenotype assessment, and integration of functional genomic data to further refine subphenotype definitions and enhance the translational potential of genetic discoveries for personalized endometriosis management.

Benchmarking Performance: Validation Frameworks and Comparative Analyses

Within endometriosis research, accurate case definition is a fundamental prerequisite for valid genetic and epidemiological studies. The development of polygenic risk scores (PRS) for disease subphenotypes is particularly sensitive to how endometriosis cohorts are ascertained. This document outlines application notes and protocols for validating two primary cohort definitions: those based on surgical confirmation (the clinical gold standard) and those derived from administrative health data (e.g., ICD codes).

A critical understanding of the operating characteristics—including sensitivity, specificity, and agreement metrics—between these two definitions is essential. It ensures that PRS models are trained on reliably classified phenotypes, thereby enhancing the predictive accuracy and clinical utility of the resulting scores for specific endometriosis manifestations [66] [11].

Comparative Data Analysis

The table below summarizes key validation metrics from recent studies comparing surgically confirmed endometriosis with cases identified through administrative health data.

Table 1: Validation Metrics of Administrative Data Against Surgical Confirmation for Endometriosis

Endometriosis Phenotype	Sensitivity (Range)	Specificity (Range)	Agreement (Kappa Statistic)	Key Findings and Implications for PRS
Overall Endometriosis [66]	0.86 - 0.88	0.83 - 0.87	0.65 - 0.74 (Substantial)	Administrative data shows high validity for etiologic studies of general endometriosis risk. Suitable for initial PRS development.
Superficial Peritoneal Disease [66]	~0.86	~0.83	~0.65 (Substantial)	Reasonably well-captured, allowing for genetic studies of this common subphenotype.
Ovarian Endometrioma [66]	~0.82	~0.92	~0.58 (Moderate)	High specificity is valuable for case-control genetic studies focusing on ovarian disease.
Deep Infiltrating Endometriosis [66]	~0.12 (Very Low)	~0.99 (Very High)	~0.17 (Slight)	Poorly captured by codes. Low sensitivity undermines statistical power for subtype-specific PRS; high specificity is only useful for pure control selection.
Self-Reported Endometriosis [67]	Variable (Literature: 32-89%)	N/A	N/A	Concordance varies widely by population and questionnaire. Requires rigorous validation against clinical records before use in genetic studies.

Table 2: Characteristics of Cohort Types for Endometriosis PRS Research

Cohort Definition	Gold Standard Status	Primary Advantages	Primary Limitations	Recommended Use in PRS Pipeline
Surgically Confirmed	Yes	High diagnostic certainty; allows for precise subphenotyping (rASRM stage, lesion location) [66] [47].	Invasive; expensive; cohort sizes may be limited; potential selection bias towards symptomatic cases.	Ideal for discovery and training of subphenotype-specific PRS models.
Administrative Health Data (ICD Codes)	No	Large sample sizes; population-based; cost-effective for very large studies [66] [11].	Misclassification bias (see Table 1); limited clinical detail; heterogeneity in coding practices.	Best for initial testing and validation in large, independent cohorts, or for studies of broad endometriosis risk.
Self-Reported	No	Easy to collect via questionnaire; can reach very large numbers.	High potential for misclassification; recall bias; cannot distinguish subtypes [67].	Use with extreme caution; requires internal validation substudy against clinical data.

Experimental Protocols

Protocol: Validation of Administrative Data Against a Surgical Gold Standard

Objective: To quantify the agreement between endometriosis diagnoses recorded in administrative health databases (e.g., using ICD-9/10 codes) and surgically confirmed diagnoses in a cohort of individuals who underwent laparoscopy/laparotomy.

Materials: Cohort with linked surgical and administrative data (e.g., Utah Population Database, ENDO Study cohort) [66].

Procedure:

• Cohort Selection: Identify all females of reproductive age within a defined population and time period who underwent a gynecologic laparoscopy or laparotomy. Exclude individuals with a prior diagnosis of endometriosis.
• Reference Standard Ascertainment: Review surgical reports and surgeon-completed standardized forms (e.g., rASRM forms) to classify patients as endometriosis cases or controls based on visualized disease. Record subphenotypes such as superficial peritoneal, ovarian endometrioma, and deep infiltrating disease [66] [47].
• Index Test Ascertainment: From the linked administrative database, extract all ICD diagnostic codes (e.g., ICD-9 617.x, ICD-10 N80.x) for the same individuals from inpatient and ambulatory surgery records over a contemporaneous time window.
• Data Analysis:
  • Construct a 2x2 contingency table (Surgical Case vs. Surgical Control vs. ICD Code Present vs. ICD Code Absent).
  • Calculate sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV).
  • Compute the Kappa statistic (Κ) to measure agreement beyond chance. Interpret Kappa as follows: ≤0.20 (slight), 0.21-0.40 (fair), 0.41-0.60 (moderate), 0.61-0.80 (substantial), 0.81-1.00 (almost perfect) [66].

Protocol: PRS Development and Validation Across Differentially Ascertained Cohorts

Objective: To develop a PRS for endometriosis and test its performance in cohorts defined by surgical confirmation and administrative codes.

Materials: Genotyped cohorts: 1) Surgically confirmed cases and controls from a clinical referral center, 2) Cases and controls identified from a population biobank using ICD-10 codes (e.g., UK Biobank, Danish registries) [11] [68].

Procedure:

• PRS Construction:
  • Obtain summary statistics from a large, well-powered endometriosis Genome-Wide Association Study (GWAS) [69].
  • Clump SNPs to select independent, genome-wide significant variants.
  • Calculate the PRS in the target cohorts using the formula: PRS = Σ (β_i * G_i), where β_i is the effect size of the i-th risk allele from the GWAS summary statistics, and G_i is the individual's allele count (0, 1, 2) [11] [9].
• Association Analysis:
  • In the surgically confirmed cohort, use logistic regression to test the association between the PRS (per standard deviation increase) and endometriosis case-control status, adjusting for genetic principal components. Repeat for major subtypes if sample size permits [11].
  • In the administrative code cohort, perform the same logistic regression analysis using ICD-10-based case definitions (N80.1-N80.9).
• Performance Comparison:
  • Compare the Odds Ratios (OR) and p-values for the PRS association between the two cohort types.
  • Assess the discriminative accuracy using the Area Under the Curve (AUC) of the Receiver Operating Characteristic (ROC) curve in each cohort. Note that AUC is expected to be lower in administratively defined cohorts due to phenotypic misclassification [11].

Visualization of Workflows

Cohort Validation and PRS Integration Workflow

PRS Analysis Across Cohort Definitions

The Scientist's Toolkit

Table 3: Essential Research Reagent Solutions for Cohort Validation & PRS Studies

Item / Resource	Function / Application	Specific Examples / Notes
Linked Biobanks & Health Registries	Provides genotype data linked to longitudinal health records for validation and large-scale genetic studies.	Utah Population Database (UPDB) [66], UK Biobank [11] [9], Danish National Patient Registry [11].
Standardized Surgical Forms	Ensures consistent and comprehensive intraoperative data collection for precise phenotyping.	Revised ASRM (rASRM) operative form [66] [47].
Genotyping Arrays & Imputation	Provides genome-wide SNP data for PRS calculation.	Commercial arrays (e.g., Illumina Global Screening Array) followed by imputation to reference panels (e.g., 1000 Genomes).
PRS Calculation Software	Tools to compute polygenic risk scores from genotype data using external GWAS summary statistics.	PLINK1.9/2.0 [9], PRSice, LDPred2.
GWAS Summary Statistics	The base data containing SNP effect sizes and p-values used to weight SNPs in the PRS.	Publicly available from largest endometriosis GWAS meta-analyses [11] [69].
Statistical Analysis Software	Platform for performing validation statistics and genetic association analyses.	R, Python, SAS.

This application note provides a detailed examination of the performance metrics and discriminatory accuracy of polygenic risk scores (PRS) across different subtypes of endometriosis. Endometriosis is a complex gynecological disorder affecting 6-10% of reproductive-aged women, characterized by the presence of endometrial-like tissue outside the uterine cavity [70] [1]. The disease demonstrates significant heterogeneity in its clinical presentation and localization, necessitating subtype-specific diagnostic and risk assessment approaches.

The current gold standard for diagnosis—laparoscopic surgery with histological confirmation—presents significant clinical challenges, with diagnostic delays typically ranging from 7 to 11 years [9] [1]. Polygenic risk scores, which aggregate the effects of multiple genetic risk variants into a single measure, offer promising avenues for non-invasive risk stratification and early detection. However, their performance varies considerably across different endometriosis subtypes, necessitating careful evaluation of their discriminatory accuracy for each major disease manifestation.

This document provides researchers and drug development professionals with comprehensive experimental protocols, performance metrics, and methodological frameworks for assessing PRS utility across the endometriosis subtype spectrum, with particular focus on ovarian, infiltrating, peritoneal, and deep infiltrating disease variants.

Quantitative Performance Metrics Across Subtypes

Polygenic Risk Score Performance

Table 1: PRS Discriminatory Performance Across Endometriosis Subtypes

Endometriosis Subtype	Cohort	Odds Ratio (OR) per SD PRS Increase	P-value	Sample Size (Cases/Controls)
Overall Endometriosis	Danish Combined	1.57	2.5×10−11	389/664
Overall Endometriosis	UK Biobank	1.28	<2.2×10−16	2,967/256,222
Ovarian (N80.1)	Danish Combined	1.72	6.7×10−5	75/NR
Infiltrating (N80.4, N80.5)	Danish Combined	1.66	2.7×10−9	210/NR
Peritoneal (N80.2, N80.3)	Danish Combined	1.51	2.6×10−3	60/NR
Superficial	Utah ENDO Study	-	-	143/412
Ovarian Endometriomas	Utah ENDO Study	-	-	38/412
Deep Infiltrating	Utah ENDO Study	-	-	58/412

The PRS demonstrates varying discriminatory ability across endometriosis subtypes, with the strongest association observed for ovarian endometriosis (OR=1.72) and the weakest for peritoneal disease (OR=1.51) [11]. Notably, the discriminative accuracy is not yet sufficient for standalone clinical utility but may add significant value when combined with classical clinical risk factors and symptoms [11].

Diagnostic Validation Metrics by Subtype

Table 2: Validation Metrics for Administrative Health Data vs. Surgical Diagnosis

Endometriosis Subtype	Sensitivity	Specificity	Kappa (Κ) Agreement
Overall Endometriosis	0.88	0.87	0.74
Superficial Endometriosis	0.86	0.83	0.65
Ovarian Endometriomas	0.82	0.92	0.58
Deep Infiltrating Endometriosis	0.12	0.99	0.17

Deep infiltrating endometriosis shows notably low sensitivity (0.12) in administrative health data, indicating this subtype is not reliably annotated in healthcare records and may require specialized detection approaches [66]. This has significant implications for PRS validation studies that rely on diagnostically coded cohorts.

Subtype Prevalence and Genetic Epidemiology

Global Prevalence Estimates

Table 3: Global Prevalence of Endometriosis and Adenomyosis Subtypes

Condition/Subtype	Population	Prevalence % (95% CI)	Number of Studies
Adenomyosis (focal)	General	17% (7-30)	59
Adenomyosis (diffuse)	General	15% (9-23)	59
Peritoneal Endometriosis	General	6% (1-15)	68
Ovarian Endometriosis	General	13% (5-24)	68
Deep Endometriosis	General	10% (2-24)	68
Endometriosis (any)	Infertile women	38% (25-51)	68
Adenomyosis (any)	Infertile women	31% (10-58)	59

Recent systematic reviews indicate that endometriosis affects approximately 38% of women experiencing infertility, with ovarian endometriosis being the most prevalent specific subtype (13%) in the general population [71]. The PRS for endometriosis shows no significant association with adenomyosis, suggesting these conditions are driven by different genetic risk variants despite shared clinical features [11].

Experimental Protocols for PRS Development and Validation

Core Protocol: PRS Calculation and Validation

Protocol 1: Polygenic Risk Score Development for Endometriosis Subtyping

4.1.1 Study Design and Cohort Identification

• Case Ascertainment: Identify endometriosis cases through surgical confirmation (laparoscopy/laparotomy) with histological examination or through validated administrative health data (ICD-10 codes N80.1-N80.9) [11] [66].
• Subtype Classification: Categorize cases into major subtypes: ovarian (N80.1), infiltrating (N80.4, N80.5), peritoneal (N80.2, N80.3), and other (N80.6, N80.8, N80.9) [11].
• Control Selection: Select age-matched controls without endometriosis diagnosis from the same population base. Exclude individuals with adenomyosis-only diagnoses (N80.0) as this represents a distinct disease entity [11].
• Sample Size Considerations: For subtype analyses, ensure adequate statistical power by including minimum sample sizes of 75 cases per subtype based on Danish cohort findings [11].

4.1.2 Genotyping and Quality Control

• Genotyping Platform: Utilize high-density genotyping arrays (e.g., Illumina Global Screening Array) [29].
• Quality Control Filters:
  • Sample-level: Exclude samples with ≥15% missing rates, ≥5% heterozygosity rate, sex discrepancies, or relatedness (PI-HAT > 0.1875) [29].
  • SNP-level: Remove markers with call rates <95%, Hardy-Weinberg equilibrium P<1×10^-5, or minor allele frequency <1% [29].
• Imputation: Perform genotype imputation using reference panels (e.g., TOPMed Version R2 on GRC38) with INFO score threshold ≥0.80 for retaining high-quality imputed variants [29].

4.1.3 PRS Calculation

• Variant Selection: Extract effect sizes (beta coefficients) and P-values from large-scale endometriosis GWAS meta-analyses (e.g., Sapkoto et al. 2017 with 14,926 cases and 189,715 controls) [9].
• Score Generation: Calculate PRS using PLINK software's score function with both unweighted (risk allele count) and weighted (effect size weighted) approaches [9] [29].

• Standardization: Convert PRS to Z-scores within each cohort to facilitate comparison across studies [9].

4.1.4 Statistical Analysis

• Association Testing: Perform logistic regression between standardized PRS and endometriosis case-control status, adjusted for principal components (typically 4-10 PCs) to control for population stratification [11] [29].
• Subtype Analysis: Conduct multinomial regression or subtype-specific case-control analyses to evaluate PRS discrimination across endometriosis subtypes [11].
• Performance Metrics: Calculate odds ratios (OR) per standard deviation increase in PRS, area under the receiver operating characteristic curve (AUC), sensitivity, specificity, and Nagelkerke's R² for variance explained [11].

Protocol 2: Validation Against Surgical Standards

4.2.1 Surgical Confirmation Protocol

• Visual Inspection: Systematic examination of pelvic cavity including uterosacral ligaments, pouch of Douglas, ovarian surfaces, and peritoneal surfaces [66].
• Lesion Documentation: Record lesion location, size, appearance (red, black, white), and depth of infiltration using standardized operative forms (rASRM) [66].
• Histological Confirmation: Obtain biopsy specimens for histological verification of endometrial glands and/or stroma in ectopic locations [66].

4.2.2 Administrative Data Validation

• Data Linkage: Link cohort data to comprehensive healthcare databases (e.g., Utah Population Database) containing inpatient, outpatient, and electronic health record data [66].
• Code Mapping: Map ICD-9/ICD-10 codes to specific endometriosis subtypes (see Table 1 for ICD-10 mappings) [11].
• Validation Metrics: Calculate sensitivity, specificity, positive predictive value, and Kappa statistics comparing administrative codes to surgical confirmation [66].

Signaling Pathways and Workflow Diagrams

PRS Analysis Workflow

Figure 1: PRS Development and Validation Workflow for Endometriosis Subtypes

Endometriosis Subtype Pathobiology

Figure 2: Pathophysiological Pathways in Endometriosis Subtypes

The Scientist's Toolkit: Research Reagent Solutions

Table 4: Essential Research Reagents and Platforms for Endometriosis PRS Studies

Category	Specific Product/Platform	Application in Endometriosis PRS Research
Genotyping Platforms	Illumina Global Screening Array	High-density genotyping for GWAS and PRS derivation [29]
Imputation Resources	TOPMed Imputation Server (Version R2)	Genotype imputation using diverse reference panels [29]
Analysis Software	PLINK (v1.9/v2.0)	PRS calculation, basic QC, and association testing [9] [29]
Analysis Software	FlashPCA	Principal component analysis for population stratification control [29]
Analysis Software	METAL	GWAS meta-analysis for variant effect size estimation [9]
Biomarker Assays	Proseek Multiplex Inflammation I	Inflammation panel (92 proteins) for subtype characterization [29]
Biomarker Assays	Olink Target 96 Platform	High-sensitivity protein biomarker detection [29]
Cell Type Characterization	CIBERSORTx Algorithm	Deconvolution of bulk transcriptomic data to estimate cell type proportions [70]
Single-Cell Reference	Marečková et al. Endometriosis Atlas	Reference scRNA-seq data for cell type annotation [70]

Discussion and Research Implications

The discriminatory accuracy of polygenic risk scores across endometriosis subtypes demonstrates significant variability, with the strongest associations observed for ovarian and infiltrating subtypes compared to peritoneal disease [11]. This heterogeneity likely reflects underlying differences in the genetic architecture and pathophysiological mechanisms driving distinct disease manifestations.

Notably, the combination of PRS with classical clinical risk factors and symptoms represents a promising approach for risk stratification tools [11]. However, researchers must account for the substantial differences in validity of subtype classification across data sources, particularly the poor sensitivity of administrative health data for deep infiltrating disease [66].

Future research directions should include the development of subtype-specific PRS using larger GWAS datasets with well-characterized surgical phenotypes, integration of multi-omics data to enhance predictive power, and investigation of gene-environment interactions across different endometriosis manifestations. The emerging understanding of cellular heterogeneity in endometriosis, particularly the role of MUC5B+ epithelial cells and dStromal late mesenchymal cells, provides new opportunities for refining subtype classification and understanding the biological mechanisms underlying genetic risk [70].

Furthermore, the association between genetic liability to endometriosis and hormonal factors, particularly the causal relationship with lower testosterone levels identified through Mendelian randomization approaches, suggests promising avenues for integrating endocrine biomarkers with genetic risk profiles for improved subtype discrimination [9].

Endometriosis, a complex gynecological disorder affecting approximately 10% of reproductive-aged women, presents significant diagnostic challenges, with average delays of 7-10 years between symptom onset and definitive diagnosis [72] [73]. The gold standard for diagnosis remains laparoscopic surgery with histological confirmation, an invasive approach with inherent risks and limited patient acceptability [73]. This application note provides a comprehensive comparison between emerging polygenic risk score (PRS) methodologies and established diagnostic markers for endometriosis, contextualized within research frameworks for subphenotype investigation and therapeutic development.

Performance Comparison: PRS vs. Traditional Biomarkers

Table 1: Comparative performance metrics of PRS versus traditional diagnostic markers for endometriosis

Parameter	Polygenic Risk Score (PRS)	Traditional Biomarker CA-125	Laparoscopy (Gold Standard)
Predictive Area (AUC)	0.744 for severe endometriosis (ML model) [47]	Limited standalone diagnostic value; often elevated but nonspecific	Not applicable (definitive diagnosis)
Odds Ratio (OR)	OR=1.57-1.72 per SD increase for various subtypes [21]	Not applicable	Not applicable
Sensitivity	Varies by model and population	36-84% (high variability) [47]	High (visual confirmation)
Specificity	Varies by model and population	41-95% (high variability) [47]	High (histological confirmation)
Sample Requirement	DNA from blood or saliva	Serum	Tissue biopsy
Key Advantages	Captures genetic predisposition; applicable pre-symptomatically; quantifiable risk stratification	Minimally invasive; low cost	Definitive diagnosis; allows concurrent treatment
Key Limitations	Currently limited predictive power as standalone tool; population-specific performance	Poor specificity; influenced by menstrual cycle, pregnancy, other pathologies	Invasive procedure; surgical risks; cost

Experimental Protocols for PRS Development and Validation

Genome-Wide Association Study (GWAS) Meta-Analysis for PRS Weightings

Purpose: To derive SNP effect sizes for PRS calculation through large-scale genetic association studies.

Detailed Protocol:

• Cohort Collection: Assemble summary statistics from multiple European cohorts (e.g., Sapkota et al. 2017 meta-analysis: 14,926 cases; 189,715 controls) and FinnGen Release 8 (13,456 cases, 100,663 controls) [9].
• Meta-Analysis: Conduct fixed-effects meta-analysis using METAL software with genomic control applied to each cohort.
• Effect Size Adjustment: Apply Bayesian methods (SBayesR as implemented in GCTB 2.02) with default settings, excluding the MHC region and imputing sample size where necessary.
• PRS Calculation: Use plink1.9's score function to calculate PRS in target cohorts (e.g., UK Biobank), converting to z-scores for phenome-wide association studies (PheWAS).
• Association Testing: Conduct PRS-PheWAS using logistic regression for binary traits and linear regression for continuous biomarkers, adjusting for principal components and age.

Traditional Biomarker Assay Protocol (CA-125)

Purpose: To quantify CA-125 levels in serum for endometriosis assessment.

Detailed Protocol:

• Sample Collection: Draw peripheral blood samples from participants following standardized venipuncture procedures.
• Sample Processing: Centrifuge blood samples at 1300-2000 × g for 10 minutes to separate serum; aliquot and store at -80°C until analysis.
• Immunoassay: Utilize electrochemiluminescence immunoassay (ECLIA) technology with ruthenium derivatives for detection.
• Quantification: Measure chemiluminescence signals and interpolate from standard curve generated with calibrators of known concentration.
• Interpretation: Apply established cutoff values (typically 35 U/mL), recognizing limited specificity for endometriosis.

Endometriosis Subphenotyping Protocol

Purpose: To classify endometriosis molecular subtypes for stratified genetic analysis.

Detailed Protocol:

• Tissue Collection: Obtain ectopic endometriotic lesions during laparoscopic surgery with patient consent and ethical approval.
• RNA Extraction: Isolate total RNA from frozen tissue samples using column-based purification methods.
• Transcriptomic Profiling: Conduct microarray or RNA-seq analysis (Illumina platforms) following standard protocols.
• Molecular Subtyping: Perform unsupervised hierarchical clustering using ConsensusClusterPlus package with settings: maxK=10, reps=10,000, pItem=0.8, pFeature=1, clusterAlg="km", distance="Euclidean" [74].
• Validation: Validate subtypes in independent datasets (GSE25628, E-MTAB-694, GSE23339) after batch effect removal using ComBat function from SVA package.

Signaling Pathways and Workflow Visualization

Diagram 1: Integrated workflow for endometriosis diagnostics combining PRS, traditional biomarkers, and clinical assessment

Diagram 2: Molecular pathways in endometriosis showing potential intervention points for PRS and biomarker applications

The Scientist's Toolkit: Research Reagent Solutions

Table 2: Essential research reagents and computational tools for endometriosis biomarker studies

Category	Specific Tool/Reagent	Application in Endometriosis Research	Key Features
Genotyping Arrays	Illumina Global Screening Array [29]	PRS variant genotyping	High-throughput SNP coverage
Imputation Reference	TOPMed Version R2 [29]	Genotype imputation	Diverse population representation
PRS Software	plink1.9 [9], SBayesR [9], GCTB 2.02 [9]	PRS calculation and weighting	Bayesian approaches for improved prediction
Biomarker Assays	Proseek Multiplex Inflammation I [29]	Inflammatory protein profiling	92 inflammation-related proteins
Immunoassays	Electro Chemi Luminescence Immunoassay (ECLI) [29]	Autoantibody detection (e.g., TRAb)	High sensitivity detection
Transcriptomic Tools	Illumina next Seq NGS technology [75]	RNA-seq for molecular subtyping	High-throughput gene expression
Clustering Algorithms	ConsensusClusterPlus [74]	Molecular subtype identification	Unsupervised pattern discovery
Machine Learning	randomForest, XGBoost, LASSO [47]	Predictive model development	Feature selection and classification

Discussion and Future Directions

The integration of PRS with traditional diagnostic markers represents a promising avenue for advancing endometriosis research and clinical management. Current evidence demonstrates that PRS captures a distinct dimension of endometriosis risk—genetic predisposition—that complements the pathophysiological information provided by traditional biomarkers. Notably, PRS shows association with all endometriosis subtypes (ovarian: OR=1.72, infiltrating: OR=1.66, peritoneal: OR=1.51) [21], suggesting broad applicability across disease manifestations.

Critical research gaps remain in optimizing PRS for diverse populations, understanding the genetic factors underlying disease subphenotypes, and integrating multimodal data sources. The identification of distinct molecular subtypes (stroma-enriched S1 and immune-enriched S2) with differential responses to hormone therapy [74] highlights the potential for PRS to guide personalized treatment approaches. Future studies should focus on developing subphenotype-specific PRS models and validating their utility in prospective clinical cohorts.

For researchers and drug development professionals, the protocols and frameworks presented herein provide a foundation for advancing precision medicine approaches in endometriosis, potentially reducing diagnostic delays and improving therapeutic outcomes for this complex condition.

The integration of polygenic risk scores (PRS) into endometriosis care represents a paradigm shift with the potential to redefine early detection and intervention strategies for this complex gynecological condition. Endometriosis, affecting an estimated 10% of women of reproductive age, is characterized by a substantial diagnostic delay of 6-11 years, during which disease progression and pain sensitization may occur [29] [76]. Clinical utility assessment provides a critical framework for evaluating how PRS—as a quantitative measure of genetic susceptibility—can impact patient outcomes and healthcare efficiency when applied to endometriosis subphenotypes. This protocol outlines comprehensive methodologies for establishing the clinical utility of PRS in expediting diagnosis, personalizing interventions, and ultimately improving quality of life for affected individuals.

Current Diagnostic Challenges in Endometriosis

The clinical landscape for endometriosis diagnosis remains challenging due to non-specific symptoms, the invasiveness of definitive laparoscopic diagnosis, and the absence of reliable non-invasive biomarkers [77] [76]. This diagnostic dilemma creates significant barriers to early intervention:

• Systemic Delays: Patients typically navigate a protracted diagnostic journey of 6-11 years from symptom onset to confirmed diagnosis [76]
• Clinical Limitations: Current imaging modalities like MRI demonstrate variable sensitivity—effective for ovarian endometriosis but poor for deep infiltrating and peritoneal lesions [77]
• Biomarker Insufficiency: Serum markers such as CA125 and CA19.9 show poor reliability, with most patients presenting normal values despite active disease [77]

These limitations underscore the urgent need for innovative risk stratification tools like PRS that can identify candidates for targeted diagnostic interventions earlier in the disease course.

Polygenic Risk Scores: Technical Foundations

PRS Calculation and Quality Control

Polygenic risk scores aggregate the effects of numerous genetic variants into a single measure of genetic susceptibility [56]. The standard approach calculates PRS as the sum of risk alleles weighted by their effect sizes derived from genome-wide association studies (GWAS) [78].

Table 1: Essential Quality Control Steps for PRS Analysis

Data Component	QC Parameter	Threshold	Rationale
Base Data (GWAS)	Heritability (h²snps)	>0.05	Ensures sufficient genetic signal
	Effect allele specification	Must be clearly defined	Prevents direction errors in association
Target Data	Sample missingness	<0.02	Reduces genotyping error
	Minor allele frequency	>0.01	Filters rare variants
	Imputation quality	INFO score >0.8	Ensures reliable imputed genotypes
	Heterozygosity	P > 1×10⁻⁶	Identifies sample contamination

Critical quality control measures must be implemented to ensure PRS validity [78]:

• Base Data QC: GWAS summary statistics must demonstrate sufficient heritability (h²snp > 0.05) with clearly documented effect alleles
• Target Data QC: Genotype data requires standard GWAS quality control including checks for missingness, Hardy-Weinberg equilibrium, and population stratification
• Ancestry Considerations: PRS performance attenuates when applied across ancestral groups, necessitating population-specific calibration [56]

Methodological Workflow for PRS Development

The following diagram illustrates the standard workflow for PRS development and validation:

Quantitative Evidence for Endometriosis PRS Performance

Recent studies demonstrate the discriminative ability of PRS for endometriosis across diverse cohorts:

Table 2: Performance Metrics of Endometriosis PRS Across Studies

Cohort	Sample Size	Odds Ratio per SD	P-value	Clinical Implications
Surgically Confirmed Cases [79]	249 cases, 348 controls	1.59	2.57×10⁻⁷	Strong association with confirmed disease
Danish Twin Registry [79]	140 cases, 316 controls	1.50	0.0001	Validates genetic component
Combined Danish Cohorts [79]	389 cases, 664 controls	1.57	2.5×10⁻¹¹	Consistent effect across recruitment strategies
UK Biobank [79]	2,967 cases, 256,222 controls	1.28	<2.2×10⁻¹⁶	Confirmation in large-scale biobank
Subtype Analysis (Combined) [79]

• Ovarian: 1.72 (p=6.7×10⁻⁵)
• Infiltrating: 1.66 (p=2.7×10⁻⁹)
• Peritoneal: 1.51 (p=2.6×10⁻³) | Comprehensive risk across manifestations |

Key findings from these studies indicate:

• PRS captures increased risk for all endometriosis types rather than specific locations [79]
• The association is specific to endometriosis, with no significant association found with adenomyosis [79]
• Current discriminative accuracy, while statistically significant, remains insufficient for stand-alone clinical use [79]

Standardized Phenotyping Protocols for Subphenotype Discovery

Robust phenotyping is fundamental to PRS clinical utility assessment for endometriosis subphenotypes. The World Endometriosis Research Foundation Endometriosis Phenome and Biobanking Harmonisation Project (EPHect) has developed standardized instruments for surgical data collection [80]:

EPHect Surgical Data Forms

• Minimum Required Form (MSF): Core elements essential for all endometriosis research
• Standard Recommended Form (SSF): Comprehensive phenotyping including detailed lesion descriptions, procedural information, and sample collection metadata

Critical Phenotyping Elements

• Lesion Characteristics: Location, type (superficial, deep, ovarian), appearance, size
• Procedural Details: Surgical approach, extent of excision, potential residual disease
• Symptom Correlation: Standardized pain and quality of life metrics
• Comorbidity Assessment: Including infertility status and associated conditions

Experimental Protocols for Clinical Utility Assessment

Protocol 1: PRS Association with Clinical Subphenotypes

Objective: To evaluate the association between PRS and specific endometriosis clinical presentations.

Methodology:

• Participant Recruitment: Women with surgically confirmed endometriosis (n≥100 recommended) [78]
• Data Collection:
  • Standardized clinical phenotyping using EPHect instruments [80]
  • Symptom assessment including pain scales and gastrointestinal symptoms (e.g., VAS-IBS) [29]
  • Documentation of disease extent, locations, and previous treatments
• Genotyping and PRS Calculation:
  • DNA extraction from blood samples
  • Genotyping using Illumina Global Screening Array or equivalent
  • Quality control per established guidelines [78]
  • PRS calculation using published endometriosis GWAS effect sizes [29]
• Statistical Analysis:
  • Logistic regression evaluating association between PRS and subphenotypes
  • Adjustment for principal components to account for population stratification
  • Specificity and sensitivity calculations for predictive performance

Protocol 2: Assessing Impact on Diagnostic Timing

Objective: To determine whether PRS-guided triage reduces time to diagnosis.

Methodology:

• Study Design: Prospective cohort study comparing diagnostic intervals
• Participant Groups:
  • High-PRS group (top 20% of distribution)
  • Average-PRS group (middle 60%)
  • Low-PRS group (bottom 20%)
• Intervention: Expedited laparoscopic assessment for high-PRS symptomatic women
• Outcome Measures:
  • Primary: Time from initial presentation to surgical diagnosis
  • Secondary: Patient-reported pain outcomes, quality of life measures
• Analysis: Survival analysis of time-to-diagnosis, comparing groups

Research Reagent Solutions for PRS Studies

Table 3: Essential Research Reagents and Platforms for PRS Investigations

Reagent/Platform	Specification	Research Function	Considerations
Genotyping Array	Illumina Global Screening Array	Genome-wide variant detection	Coverage of endometriosis-associated SNPs [29]
Imputation Reference	TOPMed Panel R2 on GRCh38	Enhances variant coverage	INFO score >0.8 recommended for QC [29]
PRS Calculation Software	PLINK (v1.9+)	Clumping/thresholding method	Industry standard for PRS computation [78] [29]
Alternative PRS Tools	PRSice, LDpred	Advanced scoring methods	Bayesian approaches for improved prediction [56]
Inflammatory Protein Panel	Olink Multiplex Inflammation I	Analyzes 92 inflammatory proteins	Identifies protein correlates of PRS [29]
Quality Control Tools	FlashPCA, standard GWAS QC	Population stratification control	Essential for confounding reduction [78]

Clinical Utility Assessment Framework

The clinical utility of PRS must be evaluated through multidimensional assessment:

Key Utility Domains

• Diagnostic Impact: Potential to reduce diagnostic delay through risk-based triage [79] [76]
• Risk Stratification: Identification of women who would benefit from enhanced surveillance or early intervention [56]
• Prevention Opportunities: Targeting of high-risk individuals for preventive strategies [81]
• Personalized Management: Informing treatment selection based on genetic susceptibility profiles [56]
• Economic Evaluation: Assessment of cost-effectiveness and healthcare resource implications [82]
• Implementation Considerations: Addressing ethical frameworks and health equity across diverse populations [56]

Limitations and Future Directions

Despite promising associations, several challenges remain for clinical implementation:

• Ancestry Limitations: Current PRS are primarily derived from European ancestry cohorts, with attenuated performance in diverse populations [56]
• Subphenotype Specificity: The association between PRS and specific clinical presentations requires further investigation [29]
• Integrative Models: Maximal utility will likely require combination with clinical, biochemical, and imaging biomarkers [79] [81]
• Intervention Protocols: Evidence-based guidelines for managing high-PRS individuals need development

Future research should prioritize:

• Large-scale GWAS specifically powered for endometriosis subphenotypes
• Development of transancestral PRS with equitable performance across populations
• Randomized trials evaluating PRS-guided care pathways on diagnostic outcomes and patient satisfaction
• Integration of PRS with electronic health records for pragmatic implementation studies

Polygenic risk scores represent a promising tool for enhancing early detection and intervention timing in endometriosis. Current evidence demonstrates significant association between PRS and endometriosis risk across multiple cohorts, with potential for stratifying women based on genetic susceptibility. However, clinical implementation requires careful attention to standardized phenotyping, methodological rigor in PRS calculation, and comprehensive assessment of clinical utility across multiple domains. As research advances, PRS-guided strategies may ultimately reduce the protracted diagnostic journey that currently characterizes endometriosis, enabling earlier intervention and improved quality of life for affected individuals.

The integration of polygenic risk scores (PRS) into endometriosis care represents a promising paradigm shift towards precision medicine. Current evidence, primarily from modeling studies, indicates a positive trend toward the cost-effectiveness of PRS-based strategies. These approaches largely focus on optimizing screening programs and refining eligibility for preventive therapies [83]. However, the field faces significant challenges, including limited real-world evidence, questions concerning the generalizability of findings across diverse populations, and a need to fully account for implementation costs and long-term benefits [83]. The following analysis provides a structured overview of the economic landscape, detailed protocols for evaluation, and essential research tools to advance the cost-benefit understanding of PRS implementation for endometriosis.

Table 1: Summary of Economic Evaluation Evidence for PRS-based Approaches

Evaluation Aspect	Current Evidence Status
Overall Trend	Positive trend towards cost-effectiveness identified in systematic review [83].
Primary Applications	1. Optimization of cancer screening programs (16 out of 24 studies) [83].2. Refinement of eligibility for preventive therapies (esp. in cardiovascular and other diseases) [83].
Analysis Quality	Generally high quality among 24 included cost-utility analyses [83].
Key Methodological Limitations	Reliance on hypothetical cohorts; limited generalizability; insufficient attention to implementation costs and delivery models; focus on clinical benefits only [83].
Evidence Gaps	Limited use of real-world data; issues of population representativeness; gaps in accounting for long-term health and non-health benefits [83].

Table 2: Performance Metrics of an Endometriosis-Specific PRS Data based on a PRS derived from 14 genetic variants, validated across multiple cohorts [21] [11] [68].

Cohort	Case Definition	Odds Ratio (OR) per SD increase in PRS	P-value
Danish Clinical Cohort	Surgically confirmed	1.59	2.57 × 10-7
Danish Registry Cohort	ICD-10 codes	1.50	0.0001
UK Biobank	ICD-10 codes	1.28	< 2.2 × 10-16

Experimental and Evaluation Protocols

Protocol for a Cost-Utility Analysis of Endometriosis PRS

This protocol outlines a methodology for evaluating the long-term economic and health impacts of implementing a PRS for endometriosis risk stratification.

1. Study Design and Model Framework

• Type of Analysis: Cost-utility analysis.
• Model Structure: Develop a state-transition Markov model (microsimulation recommended) to track a hypothetical cohort of women from a young age (e.g., 18) over a lifetime time horizon.
• Perspective: Healthcare sector and societal.
• Comparator: Usual care pathway (typically based on symptom presentation).

2. Model Parameters and Data Inputs

• Clinical Effectiveness: The model should be populated with data on the PRS's ability to reclassify risk. Use parameters such as Odds Ratios (OR) from validation studies (see Table 2) [21] [11] [68]. For instance, an OR of 1.59 per standard deviation increase in PRS indicates the increased genetic risk for surgically confirmed endometriosis.
• Epidemiological Data: Incorporate natural history data for endometriosis, including prevalence, progression rates, and associated comorbidities (e.g., infertility, chronic pain, anxiety/depression) [84].
• Cost Data: Include direct medical costs (genotyping, imaging, laparoscopy, medical and surgical treatments, management of complications) and indirect costs (productivity losses). A systematic literature review is necessary to obtain these values.
• Utility Weights: Use quality-of-life weights (e.g., from EQ-5D studies) for health states like "symptomatic undiagnosed endometriosis," "diagnosed and managed," and "post-successful treatment."

3. Outcome Measures

• Primary Outcomes: Incremental Cost-Effectiveness Ratio (ICER), expressed as cost per Quality-Adjusted Life-Year (QALY) gained.
• Secondary Outcomes: Number of laparoscopies avoided, reduction in diagnostic delay, cases of severe endometriosis prevented.

4. Analysis

• Base Case Analysis: Run the model with the most plausible input values.
• Sensitivity Analysis:
  • Probabilistic Sensitivity Analysis (PSA): Vary all input parameters simultaneously over their probability distributions to generate a cost-effectiveness acceptability curve (CEAC).
  • Scenario Analysis: Test different implementation scenarios (e.g., PRS combined with other biomarkers, targeting specific high-risk subphenotypes).

The workflow for this economic evaluation is outlined below.

Protocol for a PRS-PheWAS to Identify Comorbidities and Broader Economic Impact

This protocol describes how to conduct a Phenome-Wide Association Study using an endometriosis PRS to uncover genetic correlations with comorbid conditions, which can inform a more complete assessment of the economic impact of PRS implementation [9].

1. Polygenic Risk Score Calculation

• Genetic Data: Use genome-wide genotype data from a large biobank (e.g., UK Biobank).
• PRS Generation: Calculate an endometriosis PRS for each individual using effect sizes (weights) from a large-scale, independent endometriosis GWAS. Advanced methods like SBayesR are recommended for optimal weighting [9].

2. Phenotype Data Preparation

• Phecode Mapping: Map International Classification of Diseases (ICD) codes from electronic health records to phecodes, which group related diagnoses into single phenotypes.
• Cohort Definition: Conduct analyses in three distinct cohorts to dissect pleiotropic effects:
  • Females: The primary cohort for assessing associations.
  • Males: To identify effects independent of female reproductive anatomy.
  • Females without an endometriosis diagnosis: A sensitivity analysis to find effects not dependent on the physical disease manifestation [9].

3. Statistical Analysis

• Association Testing: For each phecode, run a logistic regression model with the phecode status as the dependent variable and the PRS (converted to a Z-score) as the independent variable.
• Covariates: Adjust for age and the first 10 genetic principal components to account for population stratification.
• Significance Threshold: Apply a multiple testing correction (e.g., Bonferroni) based on the number of tested phecodes.

4. Interpretation and Economic Implications

• Traits significantly associated with the endometriosis PRS across all cohorts, especially in males, suggest shared genetic pathways that are not a consequence of endometriosis itself. This can reveal the broader health burden attributable to the genetic liability for endometriosis, which should be included in comprehensive cost-benefit models [9].

The workflow for this analysis is as follows.

The Scientist's Toolkit: Research Reagent Solutions

Table 3: Essential Research Materials and Tools for Endometriosis PRS and Economic Research

Item / Tool	Function / Application	Example / Note
Genotyping Array	Genome-wide genotyping to generate raw genetic data for PRS calculation.	Illumina Global Screening Array [29].
Imputation Reference Panel	Increases genomic coverage by predicting ungenotyped variants.	TOPMed Imputation Server (Version R2 on GRC38) [29].
GWAS Summary Statistics	Source of SNP effect sizes (weights) for PRS calculation.	Published large-scale endometriosis GWAS (e.g., GCST004549) [29].
Statistical Software (PLINK)	A core tool for genome data management, quality control, and PRS calculation.	Used for --score function to calculate PRS [9].
Bayesian Analysis Tool (GCTB)	Implements advanced methods for PRS weighting to improve predictive performance.	SBayesR method for adjusting GWAS summary statistics [9].
Phecode Catalog	Provides the mapping system to convert ICD codes into research-ready phenotype groups (phecodes).	Essential for PRS-PheWAS to define traits for association testing [9].
Health Economic Modeling Software	Platform for building and running state-transition models for cost-effectiveness analysis.	R, TreeAge, SAS, or Excel with specialized add-ins.

Conclusion

The development of polygenic risk scores for endometriosis subphenotypes represents a transformative approach to addressing the significant diagnostic delays and heterogeneity that have long challenged clinical management. Current evidence demonstrates that PRS can stratify risk for all major endometriosis subtypes, though standalone predictive power remains insufficient for direct clinical implementation. Future research must prioritize the development of subphenotype-specific PRS, integration of multi-omics data including methylation risk scores, and diversification of genetic studies across ancestral populations. For drug development, these tools offer unprecedented opportunities for patient stratification in clinical trials and identification of novel therapeutic targets based on genetic pathways. The convergence of PRS with artificial intelligence and comprehensive biomarker panels will ultimately enable the precision medicine paradigm that endometriosis patients urgently require, potentially reducing diagnostic delays and improving therapeutic outcomes through genetically-informed care pathways.

References

• 1. Potential biomarkers for early detection of endometriosis [https://link.springer.com/article/10.1007/s13353-025-01021-y]
• 2. Identification of Candidate Genes for Endometriosis in a ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12383487/]
• 3. How to: perform polygenic risk score analysis [https://frontlinegenomics.com/how-to-perform-polygenic-risk-score-analysis/]
• 4. New therapeutic targets for endometriosis predicted ... [https://www.frontiersin.org/journals/genetics/articles/10.3389/fgene.2025.1631446/full]
• 5. An extensive multi-variant deep neural network approach ... [https://www.sciencedirect.com/science/article/pii/S092523122502168X]
• 6. Identification and Validation of Novel Combinatorial ... [https://pubmed.ncbi.nlm.nih.gov/40832379/]
• 7. Regulatory Effects of Endometriosis-Associated Genetic Variants [https://pmc.ncbi.nlm.nih.gov/articles/PMC12385710/]
• 8. Genetic variants underlying risk of endometriosis [https://pmc.ncbi.nlm.nih.gov/articles/PMC4132588/]
• 9. Polygenic risk score phenome-wide association study reveals ... [https://bmcmedicine.biomedcentral.com/articles/10.1186/s12916-023-03184-z]
• 10. Meta-analysis identifies five novel loci associated with ... [https://www.nature.com/articles/ncomms15539]
• 11. Polygenic Risk Score Prediction for Endometriosis - PMC [https://pmc.ncbi.nlm.nih.gov/articles/PMC9580817/]
• 12. Multi-ancestry genome-wide association study of ... [https://pubmed.ncbi.nlm.nih.gov/40950499/]
• 13. Endometriosis-associated infertility: Multi-omics insights into ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12527851/]
• 14. Real world perspectives on endometriosis disease ... [https://www.nature.com/articles/s44294-024-00052-w]
• 15. New Study Sheds Light on Endometriosis Genetics and ... [https://frontlinegenomics.com/new-study-sheds-light-on-endometriosis-genetics-and-pain-mechanisms/]
• 16. validating the key role of cellular senescence ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12539843/]
• 17. Multi-ancestry genome-wide association study of ... [https://www.medrxiv.org/content/10.1101/2025.09.03.25335012v1]
• 18. Genetic Differences Between Endometriosis and ... [https://precisionlife.com/poster-biological-differences-between-endometriosis-and-adenomyosis?hsLang=en]
• 19. PrecisionLife at WCE2025 | Endometriosis & Women's ... [https://precisionlife.com/news-and-events/wce2025]
• 20. The inconsistent pathogenesis of endometriosis and ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12090731/]
• 21. Polygenic Risk Score Prediction for Endometriosis [https://www.frontiersin.org/journals/reproductive-health/articles/10.3389/frph.2021.793226/full]
• 22. The phenotypic and genetic association between ... [https://pubmed.ncbi.nlm.nih.gov/40262193/]
• 23. Unraveling shared diagnostic genes and cellular ... [https://www.nature.com/articles/s41598-025-93146-7]
• 24. Global endometrial DNA methylation analysis reveals ... [https://www.nature.com/articles/s42003-023-05070-z]
• 25. Beyond endometriosis GWAS: from Genomics to ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC5693320/]
• 26. Pharmacogenomics polygenic risk score: Ready or not for ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC11287822/]
• 27. Multi-ancestry genome-wide association study of ... [https://www.medrxiv.org/content/10.1101/2025.09.03.25335012v1.full-text]
• 28. Pharmacogenomics polygenic risk score for drug response ... [https://www.nature.com/articles/s41467-022-32407-9]
• 29. Applicability of polygenic risk scores in endometriosis clinical ... [https://bmcwomenshealth.biomedcentral.com/articles/10.1186/s12905-022-01788-w]
• 30. Revisiting the Risk Factors for Endometriosis: A Machine ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC9317820/]
• 31. Fast and accurate Bayesian polygenic risk modeling with ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC10183379/]
• 32. Polygenic prediction via Bayesian regression and ... [https://www.nature.com/articles/s41467-019-09718-5]
• 33. Development and evaluation of polygenic risk scores for ... [https://www.sciencedirect.com/science/article/pii/S1098360022007869]
• 34. Identification of Endometriosis Pathophysiologic-Related ... [https://www.mdpi.com/1422-0067/26/1/424/review_report]
• 35. Interaction between genetic risk and comorbid conditions ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12159439/]
• 36. Interaction between genetic risk and comorbid conditions ... [https://pubmed.ncbi.nlm.nih.gov/40369874/]
• 37. Applicability of polygenic risk scores in endometriosis ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC9166598/]
• 38. Methylation Risk Score Modelling in Endometriosis [https://www.mdpi.com/1422-0067/26/8/3760]
• 39. Endometriosis - on the intersection of modern ... [https://www.nature.com/articles/s41431-025-01977-9]
• 40. Genetic Characterization of Endometriosis Patients [https://www.mdpi.com/1422-0067/21/5/1765]
• 41. New Insights into Genetics of Endometriosis —A Comprehensive... [https://pmc.ncbi.nlm.nih.gov/articles/PMC10340419/]
• 42. Is Associated with Endometriosis Copy Number Rare Variants [https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0103968]
• 43. Regulatory Effects of Endometriosis-Associated Genetic ... [https://www.mdpi.com/2079-9721/13/8/248]
• 44. Evaluation of Polygenic Risk Models Using Multiple ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC6169739/]
• 45. Clinical implementation of polygenic risk scores [https://www.nature.com/articles/s41431-025-01931-9]
• 46. Real world perspectives on endometriosis disease ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC11802455/]
• 47. Identification and validation of a novel machine learning ... [https://www.nature.com/articles/s41598-025-96093-5]
• 48. Construction of a clinical predictive model for risk factors ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12170414/]
• 49. Leveraging effect size distributions to improve polygenic risk ... [https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1007565]
• 50. Endometriosis classification, staging and reporting systems [https://pmc.ncbi.nlm.nih.gov/articles/PMC8530712/]
• 51. Endometriosis: From Genes to Global Gaps [https://www.synlab.com/medical-newsletter/read-article/endometriosis-from-genes-to-global-gaps-1702798]
• 52. Pathway polygenic risk scores (pPRS) for the analysis of gene ... [https://journals.plos.org/plosgenetics/article?id=10.1371/journal.pgen.1011543]
• 53. Polygenic prediction of phenotypes with a neural empirical ... [https://www.medrxiv.org/content/10.1101/2025.07.21.25331918v1]
• 54. Integrating breast cancer polygenic risk scores at scale in ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12395744/]
• 55. Population-specific polygenic risk scores for people of Han ... [https://www.nature.com/articles/s41586-025-09350-y]
• 56. Polygenic risk scores: from research tools to clinical instruments [https://genomemedicine.biomedcentral.com/articles/10.1186/s13073-020-00742-5]
• 57. Polygenic risk scores: a biased prediction? - PMC [https://pmc.ncbi.nlm.nih.gov/articles/PMC6309089/]
• 58. Endometriosis-associated infertility: Multi-omics insights ... [https://www.frontiersin.org/journals/endocrinology/articles/10.3389/fendo.2025.1613334/full]
• 59. Systematic review on the DNA methylation role in endometriosis [https://clinicalepigeneticsjournal.biomedcentral.com/articles/10.1186/s13148-025-01828-w]
• 60. Methylation Risk Score Modelling in Endometriosis [https://pmc.ncbi.nlm.nih.gov/articles/PMC12027649/]
• 61. Quantification of Free Circulating DNA and Differential ... [https://www.mdpi.com/2218-273X/15/1/69]
• 62. Utilization of polygenic risk scores in drug development ... [https://pubmed.ncbi.nlm.nih.gov/40200755/]
• 63. Protocol for a case–control study investigating the clinical ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC8354274/]
• 64. Enhancing genetic association power in endometriosis ... - PMC [https://pmc.ncbi.nlm.nih.gov/articles/PMC11419171/]
• 65. Enhancing genetic association power in endometriosis ... [https://pubmed.ncbi.nlm.nih.gov/39315247/]
• 66. Validation of administrative health data for the identification ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC11788219/]
• 67. Validation of self-reported endometriosis [https://www.sciencedirect.com/science/article/pii/S0378512225004505]
• 68. Polygenic Risk Score Prediction for Endometriosis [https://www.ukbiobank.ac.uk/publications/polygenic-risk-score-prediction-for-endometriosis/]
• 69. New Insights into Genetics of Endometriosis—A ... [https://www.mdpi.com/2075-4418/13/13/2265]
• 70. Frontiers | Integrated analysis of single-cell and bulk transcriptomic... [https://www.frontiersin.org/journals/medicine/articles/10.3389/fmed.2025.1641982/full]
• 71. Global prevalence of adenomyosis and endometriosis : a systematic... [https://rbej.biomedcentral.com/articles/10.1186/s12958-025-01483-z]
• 72. Precision Therapeutic and Preventive Molecular Strategies ... [https://www.mdpi.com/1422-0067/26/16/7706]
• 73. The Impact of Evolving Endometriosis Guidelines on ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC11702737/]
• 74. Distinct subtypes of endometriosis identified based on ... [https://www.frontiersin.org/journals/immunology/articles/10.3389/fimmu.2023.1133672/full]
• 75. Prediction of genomic biomarkers for endometriosis using the ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12019090/]
• 76. Endometriosis Communities on Reddit: Quantitative Analysis [https://www.jmir.org/2025/1/e57987]
• 77. Endometriosis: A Retrospective Analysis on Diagnostic ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC8765166/]
• 78. A guide to performing Polygenic Risk Score analyses - PMC [https://pmc.ncbi.nlm.nih.gov/articles/PMC7612115/]
• 79. Prediction for Polygenic Risk Score Endometriosis [https://pubmed.ncbi.nlm.nih.gov/36303976/]
• 80. I. Surgical phenotype data collection in endometriosis research [https://pmc.ncbi.nlm.nih.gov/articles/PMC4230690/]
• 81. opportunities for Polygenic and prevention... risk score early detection [https://www.nature.com/articles/s41416-020-0959-7?error=cookies_not_supported&code=868fc576-c57f-4324-8b56-edd976b31f32]
• 82. Methods for determining clinical utility [https://www.sciencedirect.com/science/article/abs/pii/S0009912023002023]
• 83. Weighing the evidence on costs and benefits of polygenic ... [https://pubmed.ncbi.nlm.nih.gov/40513564/]
• 84. Global, regional, and national burden of endometriosis among ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12430804/]