Modelling evolution and migration of genomically instable metastatic cancer

Speaker Name: 
Andrew McPherson
Speaker Organization: 
Simon Fraser University, BC, Canada
Start Time: 
Wednesday, September 23, 2015 - 11:15am
End Time: 
Wednesday, September 23, 2015 - 12:15pm
475 Engineering 2
David Haussler

Reconstruction of the evolutionary and migration histories of a cancer promises to increase our understanding of the evolutionary dynamics governing cancer progression and relapse, allowing for the discovery of improved treatment options. Sequencing the genome of a tumour biopsy provides information about the genotypes of populations of tumour cells. Sequencing multiple biopsies from a single patient allow for the tracking of migration and expansion of clonal genotypes, and reconstruction of evolutionary and migration histories. However, correct interpretation of sequencing data necessitates the development of computational models that account for significant confounding factors. Sequenced samples are frequently a mixture of tumour clones and contaminating normal cells, complicating inference of clonal genotypes populating a single biopsy. Furthermore, loss of chromosomal segments removes encompassed nucleotide level changes from descendant genomes, complicating the use of such changes for phylogenetic inference. The aforementioned problems are particularly pronounced in genomically unstable cancers, for which progressive acquisition of both structural and copy number changes result in rapid clonal divergence.

In this talk, I will describe methods for characterizing and modelling the evolution and migration of genomically unstable tumours from tumour sequencing data. I will first describe ReMixT, a method that uses genome graphs to jointly model segment copy number changes and rearrangements present in subpopulations of tumour cells. I will briefly describe the application of the stochastic dollo phylogenetic model to account for loss of nucleotide level changes in genomically instable cancers. I will then describe the application of ReMixT and the stochastic dollo to a study of the evolutionary histories of 31 samples from 7 High Grade Serous Ovarian Cancer (HGSOvCa) patients.

Using targeted sequencing of bulk populations and single cells, we have cataloged the genotypes of major clonal populations and estimated their frequency in each sample. The inferred clone phylogenies and clonal mixtures of the 7 patients provide examples of the diverse modes of clonal evolution and migration in HGSOvCa, with mutational drift and intraperitoneal mixing contrasted with punctuated evolution of late emergent drivers and subsequent clonal expansions.