The Semantic Precision for AI Retrieval of Knowledge (SPARK) dbGaP Challenge rewards developers’ AI tools for variable normalization and study discovery to help researchers find relevant biomedical studies.
Funder: NIH Challenges
Due Dates: January 1, 2027 (Registration) | January 15, 2027 (Code upload) | January 31, 2027 (Final submissions)
Funding Amounts: $250,000 total prizes; each task awards $100,000 for first place and $25,000 for second place
Summary: Rewards AI systems that normalize dbGaP study variables or improve privacy-preserving biomedical study discovery and cohort feasibility assessment.
Key Information: Controlled-access individual-level data are prohibited, and systems must run without proprietary model APIs.
The National Library of Medicine (NLM) runs the SPARK dbGaP Challenge, co-sponsored by the NIH Chief AI Officer and Office of Data Science Strategy. It rewards AI-powered tools that help researchers discover studies of genomic and phenotypic associations in the Database of Genotypes and Phenotypes (dbGaP). The challenge addresses inconsistent study metadata and variable descriptions that hinder discovery and cross-study comparison, with the goal of making NIH/NLM resources trusted and widely used for public benefit.
Participants may address one or both engineering tasks: mapping unstructured variables to standardized biomedical concepts, or building conversational AI for study discovery and cohort feasibility assessment. Systems work from a fixed public-data snapshot. Conversational systems should reason across combinations of study characteristics and explain why studies meet investigators’ needs, rather than simply match keywords.