Abstract 3271: A machine learning approach for somatic mutation discovery

Abstract

Abstract Variability in the accuracy for somatic mutation detection may affect discovery of alterations and therapeutic management of cancer patients. To address this issue, we developed a somatic mutation discovery method based on machine learning approaches that outperformed other methods using tumor alterations that were experimentally validated (sensitivity of 97% vs 90-99%; positive predictive value of 98% vs 34-92%). Analysis of tumor-normal exome data from 1,376 TCGA samples using this method revealed concordance for 74% of mutation calls but also identified likely false positive and negative changes in TCGA data, including in clinically actionable genes. For melanoma and lung cancer patients treated with immune checkpoint inhibitors, determination of high-quality somatic mutation calls improved mutation load-based predictions of clinical outcome. Integration of high-quality mutation detection in clinical NGS analyses improved the accuracy of test results compared to other clinical sequencing analyses not using these approaches (sensitivity of 100% vs 50-97%; positive predictive value of 100% vs 9-66%). These analyses provide an approach for improved identification of tumor-specific mutations and have important implications for research and clinical management of patients with cancer. Citation Format: Derrick Wood, James White, Andrew Georgiadis, Beth Van Emburgh, Sonya Parpart-Li, Jason Mitchell, Valsamo Anagnostou, Noushin Niknafs, Rachel Karchin, Eniko Papp, Christine McCord, Peter Loverso, David Riley, Luis A. Diaz, Sian Jones, Mark Sausen, Victor E. Velculescu, Samuel Angiuoli. A machine learning approach for somatic mutation discovery [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2018; 2018 Apr 14-18; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2018;78(13 Suppl):Abstract nr 3271.