DelSIEVE: joint inference of single-nucleotide variants, somatic deletions, and cell phylogeny from single-cell DNA sequencing data

Abstract

AbstractThe swift advancements in single-cell DNA sequencing (scDNA-seq) have enabled quantitative assessment of genetic content in individual cells, allowing downstream analyses at the single-cell resolution. This technology considerably facilitates cancer research, yet its underlying power has not been fully exploited. Specifically, computational methods for variant calling and phylogenetic tree reconstruction struggle due to high coverage variance and allelic dropout. To address these issues, here we present DelSIEVE, a statistical method that directly models the inherent noise in scDNA-seq data for the inference of ingle-nucleotide variants (SNVs), somatic deletions, and cell phylogeny. In a simulation study DelSIEVE exhibits outstanding performance with respect to the identification of somatic deletions and SNVs. We apply DelSIEVE to three real datasets, where rare double mutant and somatic deletion genotypes are found in colorectal cancer samples. As expected with the more expressive model, for the triple negative breast cancer sample we identify several somatic deletions, with less single and double mutant genotypes as compared to those reported by our previous method SIEVE.