报告摘要：

Cancer is an evolutionary process which accumulates somatic alterations along the genome. A widely accepted theory of this process is stochastic Darwinian evolution with waves of selection sweeps shaped by the tumour microenvironment. This hypothesis is supported by abundant evidence of subclonal mutations (i.e. mutations shared by only a proportion of cancer cells) in whole-genome/exome studies across multiple cancer types, which suggests cancer evolution as a highly stochastic and personalised process. In contrast, studies of gene expression profiles have identified strong inter-tumour similarities that lead to robust molecular subtypes of cancers in breast, colon, and pancreas. How can individual disease progression end up with similar phenotypic traits? Does it suggest certain degrees of convergence of evolution? If so, how can we measure it? Answers to these questions require new ways of the quantifying evolution patterns in tumours. In this work, we propose to quantifying cancer evolution in terms of the underlying evolutionary dynamics within tumours rather than the presence or absence of certain mutations. We develop a deconvolution framework that simultaneously quantifies activities of neutral and non-neutral evolutionary dynamics. The model builds on distributions of estimated cancer cell fractions (i.e. the percentage of cancer cell caring a given mutation), provided by a variational mixture model. The method can handle samples with a large number of the variants (more than 2 million) while quantifying uncertainty in a Bayesian fashion. In an analysis of 11722 tumour samples (2778 whole genomes from the ICGC-PCAWG cohort and 8944 whole exomes from the TCGA cohort), we find consistent signatures of neutral and non-neutral evolutionary dynamics. Their activities measured by exposures in single tumours show substantial differences across multiple cancer types.

报告人简介:

Ke Yuan is an Assistant Professor in Computing Science at the University of Glasgow. He received a PhD from the University of Southampton in 2013 advised by Mahesan Niranjan. Till 04/2016, He was a postdoctoral research fellow at Cancer Research UK Cambridge Institute at the University of Cambridge working with Florian Markowetz. He joined the School of Computing Science at the University of Glasgow in 05/2016.