Machine learning classification of fibromyalgia from blood transcriptomes: confounder benchmarking and implications for endometriosis-associated pain

Gene expression from a PBMC RNA-seq cohort (96 fibromyalgia patients, 93 controls) discriminates FM with ROC-AUC 0.821–0.825, passing a permutation-label null (p=0.0099).

• Why it matters: The field lacks an objective FM biomarker, but prior blood transcriptomic claims may be confounded by sex effects and PBMC cell-composition differences.
• What they did: The authors reanalyzed the largest public PBMC RNA-seq FM cohort using leakage-free nested cross-validation with training-only feature selection/scaling, and benchmarked against sex and four-score composition proxies.
• The result: Sex alone reached AUC 0.727 and composition proxies reached AUC 0.71–0.76; female-only expression (AUC 0.783) did not significantly exceed composition-only (0.712; paired 95% CI -0.025 to 0.175), and enrichment signals disappeared after composition adjustment.