Abstract Background Ustekinumab, a monoclonal antibody targeting the p40 subunit of interleukin (IL)-12 and IL-23, is a promising therapeutic approach in patients with ulcerative colitis (UC), but predicting treatment response remains a challenge. In the present study, we focused to identify prognostic response markers to ustekinumab in patients with active UC, utilizing mucosal transcriptome information and machine-learning approaches. Methods We used 36 drug naive UC patients initiating treatment with ustekinumab. Colonic mucosal biopsies were obtained before treatment initiation. The gene expression analysis was performed using a microarray panel of 84 inflammation related genes. A differential gene expression analysis (DGEA), correlation analysis, and network centrality analysis on co-expression networks were performed to identify potential biomarkers. Additionally, machine learning (ML) models were employed to predict treatment response based on gene expression data. Results The genes BCL6, CXCL5, and FASLG, were among the significantly upregulated, while IL23A and IL23R were downregulated in non-responders compared to responders. The co-expression analysis revealed distinct patterns between responders and non-responders. The ML algorithms demonstrated a high predictive power, emphasizing the significance of the IL23R, IL23A, and BCL6 genes. Conclusion Through a variety of computational approaches, our study indicates the predictive power of mucosal expression data in predicting patient response to ustekinumab treatment.
Read full abstract