Genomic variants in Cancer: an algorithmic survey
  • FENS
  • Genomic variants in Cancer: an algorithmic survey

You are here

Title: Genomic variants in Cancer: an algorithmic survey

Speaker: Cenk Sahinalp

Date/Time: 24 Jun  2015, 15:15 - 15:50

Place: FENS 2019

Indiana University Bloomington & Simon Fraser University & Vancouver Prostate Centre A key challenge in cancer genomics is the identification and prioritization of genomic aberrations that potentially act as drivers of cancer. Recent work by our lab and others have produced computational methods to identify aberrant genes that can collectively influence possibly distant genes with differential expression. In particular we have developed Hit’nDrive, a combinatorial optimization method based on a ``random-walk facility location’’ (RWFL) formulation. Hit’nDrive uses "multi-hitting time", the expected minimum number of hops in a random walk originating from any aberrant gene to reach differentially expressed genes, to measure the aberrant gene’s "influence". With Hit’nDrive we can identify driver aberrations, i.e. those whose collective influence over differentially expressed genes in a set of cancer patients is as high as possible. Given only the expression profiles of network modules seeded by these potential drivers, we show how to subtype tumor samples accurately. In the second part of the talk we will focus on how intra-tumor heterogeneity influences current practices in cancer subtyping. While recent research suggests that tumor heterogeneity has clinical implications, in silico determination of the clonal subpopulations remains a challenge. We address this problem through a novel combinatorial method, named CITUP, that infers clonal populations and their frequencies while satisfying phylogenetic constraints and is able to exploit data from multiple samples. On deep sequencing data from cancer studies CITUP predicts subclonal frequencies and the underlying phylogeny with high accuracy.