Predicting Endometriosis Status and Menstrual Cycle Phase Using DNA Methylation
Endometriosis is a chronic inflammatory disease associated with pelvic pain, infertility, and delayed diagnosis. Growing evidence suggests that altered DNA methylation contributes to disease development and could serve as a biomarker for disease. We developed a leakage-safe machine learning pipeline to classify endometriosis case-control status and menstrual cycle phase using genome-wide DNA…
Endometriosis, a chronic inflammatory disease characterized by pelvic pain and infertility, often remains undiagnosed for extended periods. Recent research indicates that alterations in DNA methylation may contribute to the disease's development, potentially serving as a biomarker for the condition. Scientists have devised a machine learning pipeline capable of classifying endometriosis status and determining menstrual cycle phase using genome-wide DNA methylation data obtained from eutopic endometrial tissue.
This dataset comprised 984 samples, analyzed through the Illumina Infinium MethylationEPIC array, which mapped around 759,000 CpG sites. To account for technical variability, the data underwent SmartSVA batch correction.
The researchers employed ridge logistic regression models, trained on stratified 80/20 train-test splits, with regularization strength determined through stratified cross-validation. Feature selection methods incorporated ridge coefficient ranking, per-CpG t-tests, and univariate logistic regression with FDR correction. Model performance was assessed using label-shuffling analyses.
The study revealed exceptional performance in classifying menstrual cycle phase (mean cross-validation AUROC: 0.971, held-out test AUROC: 0.989), which correlates with genome-wide hormone-driven methylation patterns. However, endometriosis classification showed slightly lower but still meaningful performance (mean cross-validation AUROC: 0.854, held-out test AUROC: 0.875).
Notably, ridge coefficient-based feature selection revealed compact predictive sets of CpGs, suggesting that the methylation signal associated with endometriosis is dispersed across numerous loci rather than concentrated in a few highly predictive sites. Pathway enrichment analyses uncovered significant enrichment for menstrual cycle phase but limited enrichment for disease status following FDR correction, corroborating a diffuse endometriosis-associated signal.
These findings underscore the potential of ridge regression in identifying methylation patterns linked to both endometriosis and menstrual cycle phase, emphasizing the need to consider cycle-related epigenetic variations in endometrial DNA methylation studies.
Written by urgent.news from bioRxiv's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
