3017-LB: AI-Ready Multimodal Dataset Integrating Pancreatic and Immune Data for Type 1 Diabetes Research

Introduction and Objective: Type 1 diabetes (T1D) research requires linking pancreatic molecular phenotypes with immune and longitudinal data, yet these datasets remain siloed across repositories, limiting systematic and AI-driven analysis. We evaluated whether curating and harmonizing multimodal T1D datasets from major public resources into a unified, interoperable framework could enable cross-cohort data linkage and support scalable AI model development. Methods: Datasets were curated from the Human Pancreas Analysis Program (HPAP v4.0.0), ImmPort, and the TEDDY NCC1 cohort. Inclusion required donor-level data availability and interoperability with standardized molecular or longitudinal formats. Data were harmonized into interoperable structures and integrated at the donor level. A knowledge graph framework recorded data provenance and lineage, enabling traceable cross-cohort linkage and standardized integration. Integration completeness and cross-modal linkage coverage were assessed as primary outcomes. Results: The integrated resource includes 193 donors and 17 molecular modalities, encompassing more than 4,000 experiments from HPAP v4.0.0. Six ImmPort immunology studies related to T1D immune responses were harmonized. The TEDDY NCC1 cohort contributed 27 data assets with over 39,000 entries and longitudinal data from more than 800 children at risk for T1D. Cross-cohort linkage across pancreatic, immune, and longitudinal data layers enables donor-level analysis connecting pancreatic molecular states with immune and longitudinal datasets across stages of T1D progression. Knowledge graph integration confirmed complete provenance traceability across datasets. Conclusion: These results demonstrate that systematic harmonization of pancreatic and immune datasets is feasible and produces a traceable multimodal resource, enabling cross-system investigation of pancreatic-immune interactions in T1D and providing an AI-ready foundation for future scalable discovery in diabetes research. Disclosure K. Liu: None. X. Luo: None. D. Leng: None. H.T. Vu: None. H. Zeng: None. A. Mohammed: None. F. Feng: None. J. Vandana: None. X.B. Bao: None. Y. Wang: None. Y. Tao: None. S. Lee: None. Y. Huang: None. M. Fayyaz: None. Z. Han: None. Z. Zhang: None. R. Mao: None. A.K. Taylor: None. D.C. Saunders: None. J. Cartailler: None. S. Johnson: None. D. Tewey: None. K.G. Young: None. W. Wang: None. S. Parker: Research Support; Current; Pfizer Inc. Consultant; Ended; Novo Nordisk. M. Brissova: None. S. Chen: Stock/Shareholder; Current; iOrganBio Inc. Stock/Shareholder; Ended; Oncobeat. J. Liu: None. Funding NIH(1OT2OD038003-01)