Multi-omics, spatial omics and artificial intelligence in chronic diseases: current applications and future directions.

Chronic diseases exhibit multiscale heterogeneity and spatial organization, yet analytical frameworks often fail to connect molecular perturbations, tissue architecture, and clinical outcomes within a unified evidence framework. Bulk and single-modality approaches can identify regulatory or cellular alterations but provide limited insight into why pathogenic programs emerge in anatomical regions, how they shape disease niches, and how they drive divergent clinical trajectories. Multi-omics partially addresses this gap by resolving interconnected layers of regulation, transcription, protein activity, and metabolism, whereas spatial omics maps these processes within preserved tissue architecture, enabling the delineation of lesion topology, interface states, and multicellular microenvironments. Artificial intelligence (AI) may further extend this framework by integrating molecular, spatial, imaging, and clinical data into disease representations that support disease stratification, the discovery of mechanistic patterns, and clinically testable predictions, provided that models are externally validated, biologically interpretable, and robust to dataset bias and domain shift. In this review, we examine how multi-omics, spatial omics, and AI can bridge the disconnect between molecular profiling, spatial pathology, and clinical translation in chronic disease research. We discuss six representative disease domains before synthesizing insights into how these approaches define molecular endotypes, map pathological niches, identify recurrent mechanistic patterns, and generate interpretable surrogate readouts for diagnosis, prognosis, and treatment stratification. We also highlight limitations, including preanalytical variability, inconsistencies across platforms, limited external generalizability, and the challenges of translating high-dimensional outputs into deployable clinical tools. Finally, we outline priorities for scalable evidence systems, mechanism-guided validation frameworks, multimodal model interpretability, and prospective clinical implementation.