Abstract
Crohn's disease (CD), a subtype of inflammatory bowel disease (IBD), causes chronic gastrointestinal tract inflammation. Thirty percent of patients do not respond to anti-tumor necrosis factor (TNF) therapy. Sialylation is involved in the pathogenesis of IBD. We aimed to identify potential biomarkers for diagnosing CD and predicting anti-TNF medication outcomes in CD. Three potential biomarkers (SERPINB2, TFPI2, and SLC9B2) were screened using bioinformatics analysis and machine learning based on sialylation-related genes. Moreover, the combined model of SERPINB2, TFPI2, and SLC9B2 showed excellent diagnostic value in both the training and validation cohorts. Importantly, a Sial-score was constructed based on the expression of SERPINB2, TFPI2, and SLC9B2. The Sial-low group showed a lower level of immune infiltration than the Sial-high group. Anti-TNF therapy was effective for 94.4% of patients in the Sial-low group but only 15.8% in the Sial-high group. The Sial-score had an outstanding ability to predict and distinguish between responders and non-responders. Our comprehensive analysis indicates that SERPINB2, TFPI2, and SLC9B2 play essential roles in pathogenesis and anti-TNF therapy resistance in CD. Furthermore, it may provide novel concepts for customizing treatment for individual patients with CD.