Fig. 2.

Generative model for variant allele counts A from DNA sequencing data of a tumor. A latent (unobserved) clone tree T generates m samples, each consisting of mixtures of cells with different mutations. Each mutation is assigned to a cluster ${\mathbf{C}}_j$. A cluster i of mutations occur in fraction ${\mathbf{F}}_{i,p}$ of cells in sample p. Variant read counts A are generated for each mutation with a binomial likelihood model, given an observed total read counts D