Abstract 6607: Leveraging somatic mutations to distinguish tumor/normal cells in scRNA-seq datasets across different cancer types

Recent advancements in single-cell RNA sequencing (scRNA-seq) have significantly impacted clinical oncology. However, the accurate interpretation of scRNA-seq data remains challenging due to the limitations of current methodologies in distinguishing between different cell types within a tissue sample. Despite substantial efforts to construct comprehensive cell atlases, differentiating malignant from non-malignant cells is still a time-consuming process. Current approaches primarily rely on marker genes, gene expression signatures, and copy number alterations (CNAs), though these methods are not universally applicable to all cancer types. In our study, we explored the potential of leveraging somatic mutations to differentiate tumor and normal cells within scRNA-seq datasets from ovarian and colorectal cancer samples. We conducted a comparative analysis of the effectiveness of CNAs and single nucleotide variants (SNVs), as well as developed a supervised machine learning (ML) method leveraging their combination for improved prediction. The findings highlighted the importance of understanding the biological mechanisms underlying mutational processes for accurate tumor cell detection. Specifically, tumor cells in ovarian cancer samples exhibited a higher prevalence of CNAs, whereas colorectal cancer samples, particularly those with mismatch repair deficiency (MMRd), showed limited utility of CNAs for predicting tumor/normal cell status. In contrast, SNVs proved to be extremely valuable in identifying highly mutated tumor cells in colorectal cancer, with tumor mutational burden (TMB) serving as one of the highly effective predictors. In conclusion, our study utilizes a previously developed framework for detecting somatic mutations in scRNA-seq data for the robust and effective identification of tumor cells and their clusters in tissue samples. This framework is cancer-type agnostic and could help improve tumor cell detection in cancers with a high prevalence of CNAs. Citation Format: Cindy Kim, Eke Snoeren, Francesc Muyas, Natasha Markuzon. Leveraging somatic mutations to distinguish tumor/normal cells in scRNA-seq datasets across different cancer types [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 6607.