Abstract 7507: Towards transparent and context-specific annotation of driver and passenger mutations in cancer

Abstract Cancer arises from the accumulation of somatic genetic alterations, with only a subset acting as "drivers" that promote malignant transformation. Distinguishing between driver and passenger mutations is fundamental for both clinical and research applications, yet current in-silico variant annotation tools often fall short. Our previous evaluation of multiple algorithms revealed a concerning reliance on pre-computed, black-box models that are rarely updated and typically limited to missense mutations. These tools also lack transparency and biological context, leading to inconsistent predictions that hinder effective cancer driver identification. Building on these findings, our current study emphasizes the need for a new framework in cancer genomics. This framework should be characterized by explainable, unbiased approaches, long-term support, and the integration of recurrence data as a marker of positive selection. To address the above limitations, we leverage 20 years of manually curated COSMIC data. By systematically analysing mutation peaks or hotspots within cancer-associated genes, we uncover patterns indicative of positive selection, a hallmark of driver mutations. Our approach is enriched by integrating additional biological metrics, including dN/dS ratios, which measure selection pressure on genes, and allele frequency data from diverse human populations to account for background genetic variation. Furthermore, we incorporate gene-specific functional insights to enhance the contextual relevance of our predictions. Our integrative methodology aims to establish a robust, reproducible set of rules for annotating driver mutations, leveraging multiple layers of evidence to reduce biases and increase predictive accuracy. By prioritizing algorithmic transparency and the inclusion of biologically informed data, we lay the foundation for a more reliable classification system that adapts to emerging discoveries in cancer genomics. Ultimately, our work seeks to transform cancer mutation annotation into a reproducible and clinically actionable process, paving the way for better-targeted therapies and more precise diagnostic tools. Through this, we aspire to bridge the existing gaps in driver mutation identification and to propel cancer genomics towards its full potential in precision medicine. Citation Format: Madhumita Madhumita, Jon Teague, Philip A Beer, Zbyslaw Sondka. Towards transparent and context-specific annotation of driver and passenger mutations in 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 7507.