Abstract 4871: A new multiple feature approach for rapid and highly accurate somatic structural variation discovery from whole cancer genome sequencing

Abstract

Abstract We present SWAN, a statistical framework with implemented software package for accurate detection of genomic structural variants in next-generation sequencing (NGS) data from cancer. SWAN is designed to sensitively detect submicroscopic and lesser size genome structural variants that could possibly present at low allele frequency in the sample as may be the case in clonal subpopulations of a primary tumor. The likelihood ratio scan statistics module of SWAN collectively quantifies evidence from read-pair, hang-reads based on Marked Poisson Process models while the softclipping module additionally integrates the evidence from remapped soft-clipped reads. Using a spike-in dataset, we compared SWAN against established structural variant callers in recovering simulated homozygous and heterogeneous variants of various sizes and types. SWAN maintained high accuracy, high precision and low false positives across this broad spectrum and outperformed all the other callers. In additional benchmark with golden standard real data from sample NA12878, SWAN again showed the best-balanced accuracy among all callers and was 30% more sensitivity in detecting validated deletions than the next runner-up caller LUMPY in 50x coverage. Even at 5x coverage, SWAN was able to find 40% of the validated deletions while LUMPY did less than 20%. As an application to whole cancer genome sequencing, we analyzed a matched sequenced colorectal adenocarcinoma sample and found 39 somatic deletions, insertions and duplications, some of which could potentially drive the cancer progression. SWAN is quite efficient in memory utilization and speed, and can be used for large-scale genome-wide structural variant genotyping and discovery based on NGS data. Citation Format: Li C. Xia, John Bell, Jiamin Chen, Nancy R. Zhang, Hanlee P. Ji. A new multiple feature approach for rapid and highly accurate somatic structural variation discovery from whole cancer genome sequencing. [abstract]. In: Proceedings of the 106th Annual Meeting of the American Association for Cancer Research; 2015 Apr 18-22; Philadelphia, PA. Philadelphia (PA): AACR; Cancer Res 2015;75(15 Suppl):Abstract nr 4871. doi:10.1158/1538-7445.AM2015-4871