Transcriptomic signatures reveal distinct inflammatory pathways and metabolic dysfunction in inflammatory bowel disease

K-Dense, Brian Laffin, Vinayak Agarwal

Keywords: Transcriptomics, AI, Agentic AI, K-Dense, Bioinformatics

Abstract: Inflammatory bowel disease (IBD) represents a complex group of chronic inflammatory disorders with heterogeneous clinical presentations and variable therapeutic responses, necessitating molecular approaches to understand disease mechanisms and improve patient outcomes. We conducted a comprehensive transcriptomic analysis using bulk RNA sequencing data from the ARCHS4 database to characterize gene expression patterns across IBD subtypes and tissue states. Our analysis encompassed differential expression profiling between IBD patients and healthy controls, comparative analysis of ulcerative colitis (UC) versus Crohn's disease (CD), and examination of inflamed versus non-inflamed tissue states within patients. The results revealed profound transcriptional dysregulation in IBD, with 3,706 differentially expressed genes identified between inflamed IBD epithelium and healthy controls, demonstrating fold changes spanning -15 to +15 and statistical significance reaching p-values of 10\textsuperscript{-80}. While UC and CD shared substantial molecular overlap with 1,050 common differentially expressed genes, distinct signatures emerged, including preferential IL23A upregulation in UC and enhanced IFN-$\gamma$-inducible TH1 processes in CD. Pathway enrichment analysis consistently identified IL-17 signaling as the most significantly activated pathway across comparisons, accompanied by robust neutrophil chemotaxis and antimicrobial response signatures. Notably, inflamed tissues demonstrated coordinated suppression of metabolic pathways, particularly affecting lipid metabolism, cholesterol absorption, and bile secretion, indicating fundamental metabolic reprogramming during active inflammation. Machine learning approaches utilizing these transcriptomic signatures achieved diagnostic accuracies exceeding 98\% for IBD classification and successfully predicted treatment responses across multiple therapeutic modalities. These findings establish transcriptomic profiling as a powerful tool for IBD diagnosis, prognosis, and therapeutic selection, providing a molecular foundation for precision medicine approaches that could transform clinical management by enabling personalized treatment strategies based on individual molecular profiles rather than traditional clinical classifications alone.

Submission Number: 324