Application of spatial omics, multi-omics, and artificial intelligence in inflammatory bowel disease: Current applications and emerging opportunities

Figure 1 Illustration of a framework that integrates multi-omics, spatial omics, and artificial intelligence for the study of inflammatory bowel disease. Panel (1): Depicts the intestinal mucosa as a spatially organized ecosystem, where disruptions in epithelial integrity, localization of immune cells, and interactions between the host and microbiota lead to disease-specific inflammatory environments in inflammatory bowel disease (IBD); (2) The acquisition of multimodal data merges clinical imaging with spatial omics and traditional multi-omics techniques to capture molecular and cellular changes across various anatomical levels; (3) Artificial intelligence facilitates the integration and spatial examination of these diverse datasets, revealing intricate interactions between the host and microbiome, as well as cellular networks, and identifying predictive features of the disease; and (4) This comprehensive framework advances clinical application by fostering precision medicine approaches for IBD. IBD: Inflammatory bowel disease; AI: Artificial intelligence; MALDI: Matrix assisted laser desorption/ionization.

Figure 2 Schematic representation of the challenges that limit the translation of omics discoveries into clinically actionable precision medicine in inflammatory bowel disease. Biological heterogeneity forms the foundational layer of complexity. Multimodal data acquisition introduces additional challenges, including sampling bias, incomplete modality overlap, cross-platform variability, and multi-scale resolution mismatches. Integration of high-dimensional datasets creates a computational bottleneck driven by dimensionality overload and signal-to-noise imbalance. Artificial intelligence models face challenges related to overfitting, limited validation, and interpretability. Together, these challenges delay the realization of precision medicine outputs such as patient stratification, therapy selection, fibrosis prediction, and cancer risk mapping. A real-world data feedback loop, incorporating prospective validation and iterative recalibration, is essential to overcome some translational constraints.