PyClone: statistical inference of clonal population structure in cancer

Abstract

We introduce a novel statistical method, PyClone, for inference of clonal population structures in cancers. PyClone is a Bayesian clustering method for grouping sets of deeply sequenced somatic mutations into putative clonal clusters while estimating their cellular prevalences and accounting for allelic imbalances introduced by segmental copy number changes and normal cell contamination. Single cell sequencing validation demonstrates that PyClone infers accurate clustering of mutations that co-occur in individual cells.Human cancer progresses under Darwinian evolution where (epi)genetic variation alters molecular phenotypes in individual cells 1 . Consequently, tumours at diagnosis often consist of multiple, genotypically distinct cell populations ( Supplementary Fig. 1) 2 . These populations, referred to as clones, are related through a phylogeny and act as substrates for selection in tumour micro-environments or with therapeutic intervention 2, 3 . The prevalence of a particular clone measured over time or in anatomic space is a reflection of its growth and proliferative fitness. Thus, ascertaining the dynamic prevalence of clones can identify precise genetic determinants of phenotypes such as acquisition of metastatic potential or chemotherapeutic resistance.In this contribution, we provide a statistical model for analysis of deeply sequenced (coverage >1000x) mutations to identify and quantify clonal populations in tumours, which extends to modelling mutations measured in multiple samples from the same patient. Our approach uses the measurement of allelic prevalence to estimate the proportion of tumour ♣

Assessing intratumor heterogeneity and tracking longitudinal and spatial clonal evolutionary history by next-generation sequencing

“…For example, SciClone (16) clusters the VAFs of SNAs in copy-number neutral and LOH-free portions of the genome using a Bayesian beta mixture model. Pyclone (15) is an extension of SciClone that adds prior information elicited from copy number estimates obtained from either genotyping arrays or WGS to its Bayesian nonparametric clustering method. Neither SciClone (16) nor Pyclone (15) infers the phylogenetic relationship between subclones.…”

“…The existing methods differ by how this compensation is done. ABSOLUTE (13), EXPANDS (14), Pyclone (15), and PhyloSub (33) assume that, when a CNA event overlaps a SNA, the point mutation resides in a region with homogeneous aneuploidy, a scenario where no subclonal CNA events are allowed. Li and Li (19) conducted a detailed analysis of the complete set of scenarios covering the possible order and phase of overlapping SNAs and CNAs in developing their software CHAT.…”

“…Multidimensional scatterplots of VAFs allow higher resolution for cluster detection than the one-dimensional histogram in the single-sample case. Recent methods, such as Pyclone (15) and SciClone (16), apply Bayesian mixture models to detect these clusters. LICHeE (23) and SCHISM (24) infer phylogeny from VAFs as an acyclic directed graph network.…”

“…For example, ABSOLUTE (13), one of the earliest methods, classifies mutations as clonal or subclonal after adjusting for the estimated purity and ploidy of the sample. Most approaches for the detection of subclonal mutations treat point mutations and copy number aberrations separately (15)(16)(17)(18). In the case of point mutations, that is, single-nucleotide alterations (SNAs) and small insertions and deletions (indels), most methods rely on mixture models for the variant allele frequency (VAF) under the assumption that mutations carried by the same set of cells have the same VAF.…”

Modeling Breast Cancer Intertumor and Intratumor Heterogeneity Using Xenografts

“…Because cancer genomes constantly evolve under selective pressure, we investigated the extent of the human-to-mouse and mouse-to-mouse clonal dynamics using PyClone (Roth et al 2014;Eirew et al 2015), a method for estimating the proportion of cells in a tumor carrying each mutation and clustering them into subclones (groups of cells with the same genotype). We observed that most of the clones remained stable after engraftment and throughout passag- ing, although in certain instances some subclones increased or decreased their prevalence or even appeared or disappeared.…”