Rebalancing the Gut: How AI Solved a 25-Year Crohn's Disease Mystery (today.ucsd.edu)

UC San Diego researchers used AI plus molecular biology to settle a 25-year debate about how the Crohn’s-risk gene NOD2 controls gut immune balance. By applying a machine‑learning classifier to thousands of macrophage gene-expression profiles from healthy and IBD-affected colon tissue, they defined a 53‑gene signature that cleanly separates inflammatory, microbe‑fighting macrophages from non‑inflammatory, tissue‑repairing macrophages. One signature gene encodes girdin; biochemical follow-up showed a specific region of NOD2 binds girdin in repair macrophages to suppress excessive inflammation, clear pathogens, and enable tissue healing. The most common Crohn’s-associated NOD2 mutation removes that girdin-binding segment, skewing the macrophage balance toward inflammation. The team validated the mechanism in mice: loss of girdin produced microbiome imbalance, small-intestine inflammation and fatal sepsis-like responses, demonstrating the functional consequence of the disrupted NOD2–girdin partnership. Publishing in the Journal of Clinical Investigation, the study illustrates how AI can extract actionable cell-state signatures from complex transcriptomic data and directly link them to molecular interactions and disease phenotype. For the AI/ML and biomedical communities, this work is significant both as a roadmap for combining classification models with mechanistic follow-up, and as a concrete target (restoring NOD2–girdin interaction) for future therapeutic strategies in Crohn’s disease.