Portal de Periódicos da CAPES

Abstract 1893: A computational method for somatic versus germline variant status determination from targeted next-generation sequencing of clinical cancer specimens without a matched normal control

2014; American Association for Cancer Research; Volume: 74; Issue: 19_Supplement Linguagem: Inglês

10.1158/1538-7445.am2014-1893

1538-7445

James Sun, Garrett M. Frampton, Kai Wang, Jeffrey S. Ross, Vincent A. Miller, Philip J. Stephens, Doron Lipson, Roman Yelensky,

Radiomics and Machine Learning in Medical Imaging

Abstract Background: A key constraint in genomic testing in oncology is that matched normal specimens are not commonly obtained in clinical practice. Thus, while most clinically relevant genomic alterations have been previously characterized and do not require normal tissue for interpretation, interpretation of novel variants whose somatic status is unknown is limited. We describe our approach to predicting somatic status of genomic alterations from tumor tissue alone in a CLIA-certified, NGS-based test that interrogates 236 cancer-related genes. Methods: For each sample, we obtain a genome-wide “CGH” profile based on coverage and allele frequencies (AF) at >3,500 SNPs, which is segmented and modeled to estimate the tumor purity (p), copy number (C), and minor allele count (M) at each segment. A variant's measured AF is compared to expectation: AFgermline = (pM+1-p)/(pC+2(1-p)) versus AFsomatic = pM/(pC+2(1-p)) and prediction is made with statistical confidence based on read depth and local variability of SNP AF. Results: For validation, we examined samples from 30 lung & colon cancer patients, where we sequenced tumors and matched-normal tissue. More broadly, we examined predictions for 17 somatic hotspot mutations and 20 common germline SNPs in 2,578 clinical cancer specimens. Finally, to assess the impact of stromal admixture, we examined 3 cell lines, which were titrated with their matched normal to 6 levels (10% to 75%). Overall, predictions were made in up to 85% of cases, with 95%-99% of variants predicted correctly: Validation studyCall rateSomatic variants predicted correctlyGermline variants predicted correctly30 matched-normal samples84% (479/567)95% (311/326)99% (151/153)2,578 clinical samples at common somatic and germline variants85% (4771/5583)96% (2556/2665)98% (2062/2106)3 cell lines with varying proportions of tumor-normal admixture83% (184/222)97% (60/62)97% (118/122) Conclusions: This method leverages deep NGS to predict variant somatic status without a matched-normal control. It supports functional prioritization and interpretation of alterations discovered on routine testing and can indicate additional work-up if germline risk variants are found. When optimized and fully validated, it may inform clinical decision making and expand treatment choices for cancer patients. Citation Format: James X. Sun, Garrett Frampton, Kai Wang, Jeffrey S. Ross, Vincent A. Miller, Philip J. Stephens, Doron Lipson, Roman Yelensky. A computational method for somatic versus germline variant status determination from targeted next-generation sequencing of clinical cancer specimens without a matched normal control. [abstract]. In: Proceedings of the 105th Annual Meeting of the American Association for Cancer Research; 2014 Apr 5-9; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2014;74(19 Suppl):Abstract nr 1893. doi:10.1158/1538-7445.AM2014-1893