Data Center for Acute to Chronic Pain Biosignatures

Project Summary The Acute to Chronic Pain Signatures (A2CPS) study is generating a rich longitudinal dataset that includes clinical, behavioral, biospecimen, and neuroimaging data aimed at identifying biomarkers that predict the transition from acute to chronic pain. While this dataset holds tremendous potential for advancing pain research, its complexity and scale pose substantial barriers to access, especially for researchers without technical or computational expertise. To address this problem, we propose to develop a domain-adapted large language model (LLM) that allows users to query the A2CPS dataset using natural language. This system will be fine- tuned using both structured metadata and unstructured documentation to ensure semantically accurate and interpretable responses. We will build a user-friendly web interface to support dynamic querying, hypothesis generation, and exploratory data analysis, with outputs including summary statistics, data descriptions, and visualizations. Validation will focus on accuracy, usability, and impact on research efficiency, particularly for non- technical users. This work will improve the accessibility and utility of the A2CPS dataset, support more equitable data-driven research, and establish a generalizable framework for LLM integration with other large-scale biomedical data resources.