Machine learning-integrated molecular subtyping reveals two biologically distinct endometriosis subtypes in the EndometDB database

Wang Y, Yiqun Wang, Cai X, Xiaozhen Cai, Yi Xu, Xu Y, Ma X, Xia Ma
Computational biology and chemistry · 2026 · vol. 126(Pt 1) , pp. 109349 · doi:10.1016/j.compbiolchem.2026.109349 · PMID:42727360 · W7204242364
Other OA: closed CC0
View on OpenAlex View on PubMed View at publisher
⚙ AI-generated summary by qwen3.7-flash, 2026-08-31 ⓘ

Machine learning analysis of the EndometDB database identified two biologically distinct molecular subtypes of endometriosis, providing a refined classification framework for understanding disease heterogeneity.

One-sentence paraphrase of the abstract; not a substitute for reading it. No clinical advice. How this works

Abstract

BACKGROUND: Endometriosis affects approximately 10% of reproductive-age women, with a diagnostic delay of 7-10 years. Despite clinical heterogeneity, current rASRM staging poorly predicts treatment outcomes. Molecular subtyping may reveal biologically meaningful patient strata that complement anatomical staging. OBJECTIVE: To identify molecular subtypes of endometriosis through integrated bioinformatics and construct machine learning (ML)-based diagnostic and prognostic models. METHODS: Gene expression data from EndometDB (GSE141549; 89 endometriosis patients, 41 controls) were processed. Differentially expressed genes (DEGs), weighted gene co-expression network analysis (WGCNA), and protein-protein interaction (PPI) networks were integrated to identify core genes. Consensus clustering defined molecular subtypes. ML diagnostic (XGBoost, ensemble) and prognostic (random survival forest) models were constructed and validated by nested cross-validation. An exploratory treatment response model was additionally evaluated. RESULTS: 823 DEGs (496 upregulated, 327 downregulated), 173 core genes, and 2 molecular subtypes were identified. Subtypes showed distinct ssGSEA pathway profiles but no statistically significant differences in rASRM score (p = 0.641) or age, suggesting molecularly-defined rather than clinically-defined heterogeneity. The ensemble diagnostic model achieved AUC= 0.892 (5-fold nested cross-validation). The random survival forest prognostic model, based on a surrogate endpoint, demonstrated an OOB concordance index of 0.751. An exploratory treatment response model performed no better than chance (AUC = 0.492). CONCLUSIONS: Integrating multi-method bioinformatics with ML identified 2 distinct endometriosis molecular subtypes and yielded internally validated diagnostic and prognostic models, providing an analytical framework and candidate molecular targets for future studies of endometriosis heterogeneity.

My notes (saved in your browser only)

Citation neighborhood

Papers in the corpus that this work cites (lower rings, blue) and that cite this one (upper rings, green). Dot size scales with the paper's in-corpus citation count — bigger dot = more influential within the endo/adeno field. Click a dot to open that paper. [ expand to 2 hops ] — adds papers reached through this work's immediate citers/citees. Heavier; up to 60 extra dots.

References (33)

Source provenance

europepmc
last seen: 2026-10-15T06:08:49.021275+00:00
openalex
last seen: 2026-10-15T06:00:39.803327+00:00
pubmed
last seen: 2026-10-15T06:02:21.945644+00:00
License: CC0 · commercial use OK