Abstract 6248: Leveraging knowledge graphs and AI to uncover premalignant disease mechanisms for early detection and prevention of lung cancer

Abstract Background: Early detection and treatment of lung cancer markedly improve patient outcomes, with a 5-year survival rate of 68-92% for stage I diagnoses compared to 0-10% for stage IV. Targeting early-stage cancer is pivotal in addressing tumour heterogeneity and therapeutic resistance, critical factors for effective treatment development. This study leverages knowledge graphs and AI to identify novel premalignant disease mechanisms at the early lung cancer stage, serving as potential biomarkers and therapeutic intervention opportunities for prevention of advanced disease. Methods: In this study, we built the Early Cancer Knowledge Graph (ECKG), an integrative framework that connects multi-omics data from both external and AstraZeneca’s internal resources, employing advanced machine learning algorithms to uncover novel disease insights and therapeutic opportunities. Our methodology involves mapping the ontological landscape of early-stage lung cancer through AstraZeneca's internal Biological Insights Knowledge Graph (BIKG) and integrating comprehensive multi-omics datasets. Utilising scalable KG build and clustering algorithms, we cluster ECKG into communities based on disease mechanism similarity and rank genes based on their significance in the disease network. Results: This approach identified multiple potential novel disease mechanisms, including >20 potential targets highlighting intervention opportunities. These targets underwent pathway enrichment analysis, Gene Set Variation Analysis (GSVA), and correlation to hallmarks of cancer at the single-cell level, enhancing our understanding of their biological significance. Key pathways enriched in early lung cancer cells include MAPK/ERK signaling (KRAS, EGFR) and mTOR signaling (mTOR, AKT1/2). Notably, genes influencing immune cell states within the tumour microenvironment (TME), such as IL-1 signaling, immune checkpoints, and immunosuppressant genes like TGFB1, were highly enriched, among others. These findings underscore the importance of the TME in early lung cancer progression. Most of these targets were supported by evidence in an extensive literature review of early lung cancer patient data. Conclusion: Future research should focus on experimentally validating these targets, identifying potential intervention strategies, and recommending therapeutic approaches, especially leveraging AstraZeneca’s early oncology assets. Furthermore, our ECKG framework could be a generalisable blueprint that is capable of expanding to other critical indications of unmet need, e.g., pancreatic and childhood cancers. Ultimately, the goal is to provide a comprehensive, unbiased assessment of the early disease landscape, facilitating strategic decision-making and shifting the focus from cancer cure to prevention. Citation Format: Cristina Riquelme-Vano, Shruti Shikhare, Mark O'Donoghue, Jake Cohen-Setton, Miguel Carmona, Andrew Reynolds, Krishna Bulusu, Katarzyna Nurzynska. Leveraging knowledge graphs and AI to uncover premalignant disease mechanisms for early detection and prevention of lung cancer [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 6248.