Accounting for sources of bias and uncertainty in copy number-based statistical deconvolution of heterogeneous tumour samples

Deconvolving heterogeneous tumour samples to identify constituent cell populations with differing copy number profiles using whole genome sequencing data is a challenging problem. Copy number calling algorithms have differential detection rates for different sizes and classes of copy number alterations. This paper describes how uncertainty in classification and differential detection rates can introduce biases in measures of clonal diversity. A simulation strategy is introduced that allows differential detection rates to be adjusted for and this process is shown to minimise bias.