AI and the World's Hard Problems · Medicine & drug discovery · Nature (computational biology) · 2026-10-20
Integrating disulfidptosis patterns and machine learning identifies a novel diagnostic signature for endometriosis
Endometriosis is a heterogeneous gynecological disorder with incompletely understood molecular mechanisms, and disulfidptosis, a recently identified form of programmed cell death, had not been systematically explored in this disease. Analyzing gene expression profiles, the researchers identified two disulfidptosis-associated molecular subtypes with distinct immune features and compared four machine-learning algorithms, with Extreme Gradient Boosting (XGB) performing best. This yielded a five-gene signature comprising BTBD2, SLC4A2, SSC5D, LPPR2, and SRGAP2C, which was supported by validation in two independent external datasets and partial RT-qPCR validation, offering a potential diagnostic tool and new insights into endometriosis mechanisms.