Exploratory baseline mucosal transcriptomics and vedolizumab response in ulcerative colitis

Currently, there are still no reliable pre-treatment biomarkers for vedolizumab in ulcerative colitis. To investigate baseline mucosal signals, we reconstructed the discovery cohort from GSE73661 and evaluated the derived panel in the blood samples from GSE191328 and the independent mucosal RNA-seq cohorts from E-MTAB-7845. The reconstructed cohort comprised 24 baseline samples, including 14 composite responders and 10 non-responders. Through repeated LASSO variable selection, the study identified an exploratory gene set consisting of ZNF689, TPRXL, ZC3H10, and MRPL55. The selection procedure yielded a mean held-out AUC of 0.611 across 18 complete nested-validation repetitions; this does not independently validate the final gene set. When the fixed gene set was projected exploratorily, the AUC for 17 blood samples was 0.458 (95% CI 0.162-0.755), and the AUC for 27 independent UC mucosal samples was 0.534 (95% CI 0.293-0.775). The apparent AUC of this gene set on the discovery set was 1.000, and 500 full-process bootstrap analyses indicated that gene selection was unstable. Among the three targeted inflammatory panels, interferon-γ showed a directional difference that was not statistically significant (mean GSVA score difference -0.425; nominal P = 0.066; adjusted P = 0.197). Expression scales and clinical endpoints differed across cohorts. The above results describe post-selection separation within the discovery cohort and do not establish reproducible predictive performance, tissue specificity, or a definitive four-gene signature.