Are Current Automated Approaches for Determining the Phylogeny of Multiple Deposits Capable of Interpreting the Complexity of Cancer Evolution?

Tumour heterogeneity is central to chemotherapy resistance and disease progression in advanced malignancy. This heterogeneity arises due to the evolution of clones within the tumour cell population; the advent of high throughput sequencing has allowed the detection of different tumour cell clones within and between primary tumours and their metastases, potentially allowing mapping of tumour evolution. Several, automated bioinformatic approaches have been devised for determining tumour phylogeny from changes in genomic copy number (CN); either by the overall similarity of genomic changes between tumour deposits or by examining the occurrence of shared breakpoints. We have compared these automated approaches with a manual determination of phylogeny based upon shared breakpoints identified from four cases of metastatic colorectal cancer consisting of between 6 and 53 deposits. We illustrate several recurrent issues identified with the use of automated systems for the determination of tumour phylogeny associated with an inability to correctly identify and interpret changes in ploidy, an inability to identify heterogeneity within tumour deposits, the masking of smaller events by larger ones, over-interpretation of convergent, but unrelated events, over calling sequencing artefacts as changes in CN, and non-calling of genuine CN changes due to low tumour cell content or low sequencing depth. We conclude that manual interpretation of bioinformatics data is still required to determine the phylogeny of metastatic cancer within an individual.