CMStatistics 2017: Start Registration
View Submission - CMStatistics
Title: Reconstructing cancer phylogenies with a small number of pairwise haplotypes and bulk tumour sequencing data Authors:  Amit Deshwar - University of Toronto (Canada) [presenting]
Abstract: PhyloSpan, a novel method for reconstructing the evolutionary history of tumours from sequencing data from one or more heterogeneous tumour samples, is presented. PhyloSpan uses data from reads (or read-pairs) that cover the genomic loci of more than one somatic mutation -- in addition to SNV allele frequencies and CNA prevalence estimates used by other methods. We show that these haplotypes can resolve ambiguities in phylogenetic reconstruction and can increase the sensitivity of reconstruction methods to subclones with similar cellular prevalences. We then considered the data from 2,778 PCAWG tumour samples. By phasing SNV pairs in haploid regions in the 775 PCAWG samples with appropriate pairs, we estimate that less than 4\% of SNV pairs, on average, are in branching lineages, strongly suggesting that the vast majority of subclones are in linear phylogenies consistent with incomplete selective sweeps. Applying PhyloSpan to PCAWG samples with pairwise haplotype data, we find numerous cases where PhyloSpan can distinguish subclones with similar cellular prevalences; can merge subclonal clusters that are incorrectly deemed separate; and can detect branching lineages undetectable with standard techniques. PhyloSpan can readily applied to long read sequencing data; but even short read sequencing data often contains informative haplotypes than can be leveraged to improve cancer phylogenies.