Abstract 3370: MCC-MELD: A diverse lung cancer registry with detailed molecular and environmental annotation

Abstract Background: Research on EGFR-mutated lung cancers has been constrained by the scarcity of large, diverse datasets integrating molecular, clinical, and environmental data. To address this gap, we built the Meyer Cancer Center Molecularly Enhanced Lung Cancer Database (MCC-MELD), which combines molecular testing results, detailed clinical information, and fine-resolution environmental data while leveraging the diversity of our catchment area. Methods: MCC serves as the cancer center for Weill Cornell Medicine, comprising three clinical sites in New York City: Manhattan, Queens, and Brooklyn. We extracted electronic health record (EHR) data for patients diagnosed with lung cancer from all sites using Epic, and from Queens (AllScripts) and Brooklyn (Cerner Millennium) sites prior to Epic implementation in 2020. We obtained tumor characteristics and treatment history from the ONCOLog internal tumor registry (2010-2020) and a manually curated lung cancer database maintained by our thoracic oncology clinic (2010-2024). We retrieved molecular testing results from an in-house pathology database (cBioPortal) and from the EHR. We used natural language processing (NLP) powered by generative artificial intelligence (AI) using Gemini-1.0-pro to abstract additional factors, including quantitative smoking data. Area-level factors - including air pollution, neighborhood socioeconomic status, and built environment - were linked to geocoded patient addresses at diagnosis. To address the challenges of measuring air pollution in densely populated urban areas, we will use a hyperlocal AI model that estimates longitudinal air pollutant levels on a 1000-by-1000-meter grid scale. Results: We identified a maximally inclusive cohort of 12, 285 patients, with 8, 754 comprising the analytic cohort (2010-2024) of confirmed lung cancer patients. The cohort included 53% female patients (compared to 44% nationally), with racial and ethnic representation of 20% Asian, 13% Black, 8% Hispanic, and 46% White patients. Lung cancer cases among nonsmokers increased from 9.9% to 20.0% over the study period. Of the cases that received in-house comprehensive genomic profiling, EGFR alterations were found in 28.7% (766 of 2, 671). Our preliminary NLP achieved high accuracy in extracting smoking status (87.1%) and smoking intensity (94.4%, measured as packs smoked per day). Discussion: MCC-MELD shows the feasibility of integrating AI-powered annotation and prediction into clinical research, providing a robust platform to study risk factors, prognostic indicators, and care patterns for EGFR-mutated lung cancers. Our planned analysis using our hyperlocal air pollution model will allow us to assess the relationship between air pollutants and EGFR-mutated lung cancer at a uniquely fine scale. We expect MCC-MELD will be a valuable resource for advancing research on lung cancer risk and outcomes in a diverse urban population. Citation Format: Luchang Cui, Juhong Lee, Sajjad Abedian, Minmeng Tang, James Solomon, Robert Crupi, Juliet Miller, H. Oliver Gao, Evan Sholle, Christine Garcia, Eunji Choi, Yiwey Shieh. MCC-MELD: A diverse lung cancer registry with detailed molecular and environmental annotation [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 3370.