Sex-stratified blood transcriptomic candidate biomarkers of multiple sclerosis identified by ensemble feature selection.

Sex-related differences in multiple sclerosis (MS), including variation in incidence, disease course, immune responses, and progression, suggest that sex influences disease-related molecular mechanisms. However, the identification of robust sex-specific molecular markers remains challenging due to data heterogeneity, high dimensionality, and limited sample size. This study aimed to identify candidate sex-specific MS biomarkers, using whole-blood gene expression data and a sex-stratified transcriptomic analysis combined with an ensemble feature selection (FS) approach. Whole-blood gene expression data from 85 untreated RRMS cases and 60 healthy controls were analyzed. The results were partially validated using two independent external datasets. The ensemble FS method and random forest recursive feature elimination were applied to identify robust candidate biomarkers. Machine learning models were evaluated using k-fold cross-validation. Female-specific, male-specific, and sex-independent biomarker sets were identified, with sex-specific signatures showing strong discriminative performance (AUC-ROC: 0.822-0.865), supporting MS molecular stratification. The probe-gene mapping was performed, and gene set enrichment and gene-drug association analyses were conducted to provide biological and therapeutic context for the most relevant MS-associated genes, considering sex differences. Several previously underexplored sex-independent genes (TMEM258, CHMP2A, NDUFB4, and TOMM22) and sex-specific candidates (e.g., CFD, SLPI, and SAMD12 in males; TMEM176B and HLA-DRB5 in females) were identified as promising candidate diagnostic biomarkers for further investigation. The identified sex-specific signatures underscore the importance of biological sex in MS biomarker discovery and support the potential of blood-based biomarkers for improving disease stratification.