Abstract Background The rising global incidence of Inflammatory Bowel Disease (IBD), encompassing Crohn’s disease (CD) and ulcerative colitis (UC), highlights the etiological role of environmental factors. A comprehensive understanding of these drivers remains elusive. This study utilizes the exposome framework to investigate the potential environmental drivers of IBD. Methods We conducted an agnostic Exposome-Wide Association Study (ExWAS) on 7,992 exposures using data from 487,258 UK Biobank participants. In both the discovery and validation datasets, we employed Cox proportional hazards models to assess the association of each exposure with incident CD and UC. After hierarchical clustering and eliminating multicollinearity, multivariable Cox models identified significant external exposure patterns. Bayesian non-negative matrix factorization (bNMF) was used to integrate genetic risk loci with internal exposures to uncover mechanistic pathways. Findings were triangulated using network analysis, LDSC, and Mendelian Randomization (MR). Finally, a biomedical knowledge graph and a Graph Retrieval-Augmented Generation (GraphRAG) framework were developed. Results The ExWAS identified 260 significant exposures for CD and 110 for UC. Key modifiable external factors included protective effects from wholemeal bread consumption and happiness with own health, and risk effects from smoking (CD), white bread, and psychosocial stress. While exposure effects were strongly correlated between CD and UC (R2 = 0.63), each disease retained a unique exposomic signature. The bNMF analysis integrating genetics and internal exposures revealed three distinct mechanistic pathways for CD (T-cell dysregulation, eosinophilic inflammation, myeloid cell suppression) and three for UC (pan-inflammatory cytokine signaling, viral sensing/autoimmunity, systemic leukocyte suppression). We integrated the evidence and constructed a novel, multi-source biomedical knowledge graph, providing a platform for clinical and public health personnel to systematically query IBD risk and protective factors. Conclusion This study provides the most comprehensive exposome atlas of IBD to date. We identified a complex web of envoronmental factors contributing to IBD risk. Furthermore, we deconstructed IBD heterogeneity into distinct biological mechanisms defined by specific genetic and internal exposure profiles. These findings provide an evidence base for prevention strategies and future validations. The novel AI-powered knowledge platform offers a dynamic resource to accelerate future research and clinical inquiry. Conflict of interest: Ma, Kejia: No conflict of interest Nie, Kai: No conflict of interest Yang, Hongxi: No conflict of interest Li, Dingpeng: No conflict of interest Wang, Xiaoyan: No conflict of interest
Ma et al. (Thu,) studied this question.