New algorithms for long-read cancer genomics

Division Of Basic Sciences - Nci

Investigators

Linked publications, trials & patents

Abstract

For the detection of somatic variation in cancer genomes, long-read sequencing is advantageous over short-read sequencing because of improved mappability and direct variant phasing. However, most current long-read methods are not developed for the analysis of tumor genomes. Here we develop a set of new long-read methods for somatic SNV, SV and CNA calling in cancer genomes, complemented by a set of openly available datasets and benchmarks to encourage new developments. First, this FY we published Severus, a breakpoint graph-based algorithm for somatic SV calling that can characterize complex multibreak SV patterns and produces haplotype-specific calls. On a comprehensive multitechnology cell line panel, Severus consistently outperforms other long-read and short-read methods in terms of SV detection F1 score. We also illustrate that compared to long-read methods, short-read sequencing systematically misses certain classes of somatic SVs, such as insertions or clustered rearrangements. We also developed Wakhan - a tool for haplotype specific copy-number alteration calling, which takes advantage of long-read connectivity to produce phased representation of complex, chromosome-scale rearrangements such as chromothripsis or seismic amplification. Wakhan is now publicly available and the manuscript will be submitted by the end of this FY. Finally, we co-developed DeepSomatic, a versatile somatic small variant caller that works with various modes of short- and long-read sequencing, which consistently outperforms other popular callers.