Novel Gene Signatures Predicting Primary Non-response to Infliximab in Ulcerative Colitis: Development and Validation Combining Random Forest With Artificial Neural Network

Background: While infliximab has revolutionized the treatment of ulcerative colitis, primary non-response is difficult to predict, which limits effective disease management. The study aimed to establish a novel genetic model to predict primary non-response to infliximab in patients with ulcerative colitis. Methods: Publicly available mucosal expression profiles of infliximab-treated ulcerative colitis patients (GSE16879, GSE12251) were utilized to identify potential predictive gene panels. The random forest algorithm and artificial neural network were applied to further screen for predictive signatures and establish a model to predict primary non-response to infliximab. Results: A total of 28 downregulated and 2 upregulated differentially expressed genes were identified as predictors. The novel model was successfully established on the basis of the molecular prognostic score system, with a significantly predictive value (AUC = 0.93), and was validated with an independent dataset GSE23597 (AUC = 0.81). Conclusion: Machine learning was used to construct a predictive model based on the molecular prognostic score system. The novel model can predict primary non-response to infliximab in patients with ulcerative colitis, which aids in clinical-decision making.