BackgroundUlcerative colitis (UC) is a chronic immune-mediated inflammation of the colorectum, for which infliximab (IFX) is currently the mainstay of treatment. However, one-third of patients with UC still fail to benefit from the IFX therapy, and early exposure to IFX impairs the efficacy of other subsequent biologics. Therefore, personalized therapeutic system is urgently needed to assist in clinical decision-making and precision treatment. MethodsFour microarray datasets of colonic biopsies from UC patients treated with IFX were obtained from the GEO database to form the Training Cohort and Validation Cohort. Differentially expressed genes (DEGs) in Training Cohort were identified and enriched for GO, KEGG and immune cell infiltration analysis. A prediction model for IFX efficacy was developed based on the LASSO and Logistic regression. The predictive accuracy of the model was verified by the Validation Cohort, and the model-genes/proteins were validated by immunohistochemistry. Gene-drug, gene-ncRNA interaction analysis were performed to identify drugs or non-coding RNAs (ncRNAs) that potentially interacted with the model-genes. Homology Modeling and Molecular Docking were conducted to filter the optimal candidate as the subsequent adjuvant or alternative for IFX in predicted non-responders. At last, the down-regulation of the key model-gene/protein CYP24A1 by the drug candidate Deferasirox was verified by Western Blot and qRT-PCR Assay based on cellular experiments. ResultsA total of 113 DEGs were identified in the Training Cohort, mainly enriched in inflammatory cell chemotaxis, migration, and response to molecules derived from intestinal microbiota. Activated pro-inflammatory innate immune cells, including neutrophils, M1 macrophages, activated dendritic cells and mast cells, were significantly enriched in colons of non-responders. The prediction model based on three model-genes (IFI44L, CYP24A1, and RGS1) exhibited strong predictive efficacy, with AUC values of 0.901 and 0.80 in the Training and Validation Cohorts, respectively. Higher expression of the three model-genes/proteins in colons of non-responders to IFX was confirmed by clinical colonic mucosal biopsies. 4 Drugs (Calcitriol, Lunacalcipol, Deferasirox, Telaprevir), 15 miRNAs and 66 corresponding lnRNAs interacting with model-genes were identified. The protein 3D structure of the key model-gene/protein (human-derived CYP24A1) was developed. Through the Molecular Docking and cellular experimental validation, Deferasirox, which significantly down-regulated both the RNA and protein expression of CYP24A1, was identified as the optimal adjuvant or alternative for IFX in predicted non-responders with UC. ConclusionThis study developed a novel prediction model for pre-assessing the efficacy of IFX in patients with UC, as the first step towards personalized therapy. Meanwhile, drugs and non-coding RNAs were provided as potential candidates to develop the next-step precise treatment for the predicted non-responders. In particular, Defeasirox appears to hold promise as an adjuvant or alternative to IFX for the optimization of UC therapy.