Computational Intelligence Methods for Predicting Treatment Outcomes in Ulcerative Colitis

Abstract
Ulcerative colitis (UC) is a chronic inflammatory bowel disease with variable responses to infliximab therapy. This study aimed to predict treatment outcomes using transcriptomic data and machine learning. Two microarray datasets (GSE12251, GSE14580) and one RNA-seq dataset (E-MTAB-7604) were analyzed after preprocessing, normalization, and feature selection. ReliefF and SelectFromModel methods were applied, followed by classification with algorithms including LightGBM, Extra Trees, and KNN. On the combined microarray data, ReliefF + LightGBM and SFM + Extra Trees achieved the best performance (F1-Score = 0.74). In contrast, the RNA-seq dataset yielded lower F1-Score, with SFM + KNN reaching F1-Score = 0.55. These results suggest that gene expression-based signatures have potential for guiding infliximab therapy decisions in UC, though performance varies across data types and validation cohorts. Keywords - Computational Intelligence, Ulcerative Colitis, Machine Learning, ,Bioinformatics, Feature Selection.