Integrated analysis of single-cell and bulk transcriptomic data reveals altered cellular composition and predictive cell types in endometriosis

This study integrated a public single-cell RNA-seq dataset with multiple bulk microarray datasets to deconvolute cell-type composition across 201 samples (105 endometriosis and 96 healthy controls), followed by differential expression, pathway enrichment, and subtype marker intersection analyses. It found five major cell types altered in endometriosis, with epithelial cells proportionally decreased while mesenchymal cells increased, alongside EMT pathway activation; among epithelial subtypes, MUC5B epithelial cells showed an increasing trend aligned with later mesenchymal subtypes and eM2 macrophages. A random forest model using estimated cell-type percentages distinguished endometriosis from healthy controls with high performance (AUC 0.932), with MUC5B cells as the top contributor, though the paper’s approach relies on deconvolution from bulk data and pooled pre-existing datasets rather than new matched specimens. This paper is centrally about endometriosis — it develops a single-cell–guided deconvolution and predictive cell-type model for endometriosis and characterizes EMT-associated cellular composition changes.

Read from the paper's body, not the abstract. Not a substitute for reading the paper. No clinical advice. How this works