Abstract LB-232: Tumor clonality detection using next generation sequencing data.

Abstract

Abstract A tumor's somatic variants, regardless of their significance in disease etiology, serve as genetic markers that capture the evolutionary history of clones and thus may be used for identifying them. We have previously identified steps along this evolutionary trajectory by discovering a broad range of variants, including exonic, intronic, and intergenic mutations, using whole genome sequencing (WGS) and by obtaining deep read counts of those variants using capture-based, targeted re-sequencing. Most of these variants, in most of the tumors analyzed, form clusters. This has revealed that most tumors are multi-clonal in composition with cellular sub-populations having distinct genomes harboring diverse somatic variants. Understanding the clonal diversity and composition of a tumor sample may have implications for clinical treatment, while the clonal evolution across tumor and relapse samples provides insight into disease progression. Here we describe our approach for inferring clones using standard single nucleotide variant (SNV) and copy number alteration (CNA) data obtained from next-generation sequencing of a normal sample and one or more diseased (e.g., tumor or relapse) samples. Variants are clustered according to their deep-read-count-derived frequencies using a variational Bayesian approach to Beta mixture modeling, through which outliers and the number of clusters are identified automatically. The method's probabilistic interpretation provides a quantifiable measure of confidence in a resulting cluster, for example, through a standard error of the mean. We show that amplifications and deletions may induce artifactual clones, thus highlighting the importance of accounting for copy number events in interpreting SNV data for inferring clones. We demonstrate the performance of our approach using published acute myeloid leukemia tumor and relapse pairs as well as unpublished multiple myeloma WGS data sets. However, the approach is applicable to solid tumors as well and has been used to cluster six-dimensional variant data from a breast tumor and five relapses originating from the same patient. Citation Format: Nathan D. Dees, Christopher A. Miller, Brian S. White, William Schierding, Ravi Vij, Michael H. Tomasson, John S. Welch, Timothy A. Graubert, Matthew J. Walter, Timothy J. Ley, John F. DiPersio, Elaine R. Mardis, Richard K. Wilson, Li Ding. Tumor clonality detection using next generation sequencing data. [abstract]. In: Proceedings of the 104th Annual Meeting of the American Association for Cancer Research; 2013 Apr 6-10; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2013;73(8 Suppl):Abstract nr LB-232. doi:10.1158/1538-7445.AM2013-LB-232