Abstract 2412: Adenine AI: Advancing germline and somatic evidence discovery for variant assessment and gene validation

Abstract Technological advancements have significantly reduced the cost of panel assays, exome, and genome sequencing, yet the classification and interpretation of germline and somatic variants for pathogenicity and oncogenicity remain challenging. Irrespective of the guidelines used, genetic and functional evidence plays a crucial role in supporting variant classification and interpretation. For instance, at least eight evidence codes (PP1, PM6, PS2, PS3, PP4, BS3, BS4, and BP3) from ACMG guidelines and another eight evidence codes (OS1, OS2, OS3, OM4, OM3, OP3, OP4, and SBS2) from joint ClinGen/CGC/VICC guidelines rely on genetic or functional evidence. Published case reports, multi-patient studies, and functional assessments are essential in strengthening or refuting variant classification and gene-disease associations in both somatic and constitutional cancer contexts.However, the ever-expanding body of scientific literature—now exceeding 40 million articles—makes finding, collecting, and curating relevant evidence a daunting task. Manual efforts by expert professionals, while accurate, are labor-intensive, costly, and inherently unscalable. These limitations impede timely variant interpretation and indirectly affect patient care and diagnosis in both germline and somatic contexts.To address these challenges, we have developed an automated pipeline leveraging artificial intelligence (AI) and workflow automation to streamline literature collection and curation. Our system continuously scans newly published evidence, evaluates relevance, categorizes it into germline or somatic contexts (e.g., single case, multi-case, or functional assessments), and extracts critical details such as gene, disease, patient demographics, cohort size, study type, interventions, variant identity, and assay type. The extracted data is normalized, mapped to ontologies, and made available through our user-facing platform, Adenine AI (accessible at https://adenine.ai).Adenine AI simplifies evidence discovery through an intuitive interface while offering powerful, documented web APIs for seamless integration with electronic health records (EHR) and variant classification/interpretation systems. Beyond ease of access, our platform enables secondary analyses, confirming and extending findings on somatic hotspot mutations while revealing significant gaps in widely used resources such as ClinVar, where evidence for many genes remains absent.This study introduces our automated pipeline and Adenine AI, available for free trial, and demonstrates its potential to improve evidence accessibility and streamline variant classification workflows, ultimately advancing both research and clinical practice. Citation Format: Ronak Y. Patel. Adenine AI: Advancing germline and somatic evidence discovery for variant assessment and gene validation [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 2412.