Generative AI-augmented transcriptomic and microbiome analysis across inflammatory and fibrotic disease states in Crohn's disease.

INTRODUCTION: Intestinal fibrosis is a major complication of Crohn's disease (CD), a subtype of inflammatory bowel disease (IBD) driven by chronic inflammation and resulting in irreversible structural damage requiring surgery. However, the molecular differences between inflammatory and fibrotic CD remain poorly defined. METHODS: Here, we developed an integrated multi-omics framework combining transcriptomics, microbiome analysis, and generative AI to characterise transcriptomic differences across non-IBD (n = 176), baseline CD (n = 187), and fibrosis CD (n = 85) tissues. Bulk and single-cell RNA-seq and 16S rRNA datasets were integrated, and machine learning identified disease-stage associated features. RESULTS: A shared set of 43 genes between baseline and fibrotic CD was organised into three modules: Module 1 (S100A8, TREM1, CXCL1) linked to innate immune activation which was upregulated in fibrosis CD; Module 2 (FABP6, MGAM, ALDOB) reflecting epithelial metabolic dysfunction which was upregulated in baseline CD; and Module 3 (CHI3L1, SAA2-SAA4, IL1RN) associated with epithelial stress and loss of barrier integrity. GSVA highlighted LCN2 and MMP3 across disease states. Microbiome analysis showed depletion of SCFA-producing genera (Faecalibacterium, Anaerostipes, Coprococcus, Ruminococcus) and enrichment of Bilophila and Bacteroides. Notably, LLM-guided augmentation improved model stability and facilitated the identification of key fibrosis-associated genes, including IL-23R, TNF-α, and TGF-β. DISCUSSION: These findings suggest that intestinal fibrosis in CD does not represent a separate molecular state, but a reconfigured inflammatory condition characterised by persistent immune activation, epithelial dysfunction, and altered host-microbiome interactions.