Project Summary This proposal outlines plans for the next generation NIDDK Information Network (dkNET), whose goal is to provide a centralized resource that connects the National Institute of Diabetes, Digestive and Kidney Diseases (NIDDK) research community to the growing number of biomedical resources (e.g., organisms, reagents, materials, protocols), data, and bioinformatics tools available to them through the multiple programs and centers established by NIDDK and the broader scientific community. dkNET has seen steady growth over the last project period and achieved significant success toward our mission, highlighted by 4 main areas of impact: 1) provided a novel Resource Information Network (RIN) that connects the NIDDK research community to the expanding universe of biomedical resources and data; 2) significantly improved the rigor and reproducibility of published research through its leadership in creating and promoting adoption of Research Resource Identifiers (RRIDs); 3) increased computational skills among the NIDDK workforce through educational programs, the dkNET Bioinformatics Pilot program, the Hypothesis Center, and the D-Challenge; and 4) brought FAIR (Findable, Accessible, Interoperable, Re-usable) data principles to DK researchers through educational materials, a webinar series, and the Summer of Data student program. dkNET tools, resources, and training materials assist researchers throughout the experimental process from hypothesis generation to supporting sound and reproducible science. Building on the success of dkNET (Resource Core) we will extend its services to the community through a new Computational Core that will bring powerful new AI/ML techniques and cloud computing resources to the NIDDK research community to fully leverage data assets cataloged by dkNET to explore and develop hypotheses. Our objective is to create a unified ML paradigm for performing a diverse range of analytical tasks. This paradigm will have the ability to process the various data types cataloged by dkNET; integrate data of multiple modalities from different sources; and incorporate domain-specific knowledge from public knowledge bases. The resulting knowledge model from this paradigm will facilitate hypothesis validation and recommendation by locating data that supports or contradicts a hypothesis/task and evaluating a given hypothesis’s likelihood.