Abstract 166: Optimizing the detection of subclonal somatic variants with VarDict

Abstract

Abstract Detection of somatic variants in tumor tissues is far from a solved problem. Heterogenous tumor materials complicate variant assessment in low frequency subclonal populations, which can play an important role in the development of drug resistance. Loss of heterozygosity, copy number variation and other ploidy changes can further change the expected distribution of variant metrics. We demonstrate improvements in the specificity and sensitivity of variant calling algorithms by optimising linear combinations of widely used filtering criteria such as allelic frequency, depth and p-value distributions. We specifically improve SNP and indels from the freely available VarDict caller (https://github.com/AstraZeneca-NGS/VarDictJava) and show that these optimisations can be generalised across a wide range of allelic frequencies. We validate our findings against synthetic and in vitro standards such as the ICGC-TCGA DREAM challenge truth set. Our proposed filtering strategy is widely applicable to other variant callers, and the updated filters for VarDict dramatically improve sensitivity and precision on low frequency variants which are crucial for our ability to reconstruct likely tumor populations from short read sequencing data and to study tumor evolution. Citation Format: Zhongwu Lai, Brad Chapman, Miika Ahdesmäki, Oliver Hofmann, Justin Johnson, Jonathan Dry. Optimizing the detection of subclonal somatic variants with VarDict. [abstract]. In: Proceedings of the 107th Annual Meeting of the American Association for Cancer Research; 2016 Apr 16-20; New Orleans, LA. Philadelphia (PA): AACR; Cancer Res 2016;76(14 Suppl):Abstract nr 166.