Few-shot Learning-driven discovery of Lutein suppresses Th1-mediated inflammation via glucose metabolism

Artificial intelligence (AI)-driven drug discovery is often hindered by the “few-shot” data bottleneck and the limited representational power of traditional two-dimensional models, challenging the accurate identification of functional molecules. In this study, we addressed these challenges by establishing a high-precision screening platform, which leverages transfer learning to recognize the key 3D molecular features of T cell inhibitors. Using this approach, we identified Lutein as a novel, specific immunomodulatory agent from a natural product library. Integrated multi-omics analyses revealed that Lutein activates peroxisome proliferator-activated receptor gamma (PPARγ), suppressing glucose uptake and glycolysis, thereby selectively inhibiting Th1 cell differentiation. In a dextran sulfate sodium (DSS)-induced mouse model of ulcerative colitis, Lutein treatment significantly restored Th1-mediated immune balance and alleviated pathological tissue damage. Our findings not only highlight the great potential of “few-shot” AI strategies that leverage transfer learning and 3D molecular features for the discovery of bioactive natural compounds, but also identify Lutein as a promising therapeutic candidate for ulcerative colitis.