Advances in Understanding Tumour Evolution through HPC
TimeTuesday, June 26th9:30am - 10am
DescriptionRecent advances in sequencing technology make it possible to study the mutational processes of cancer development and progression at an unprecedented resolution. Through single-cell sequencing it is now possible to obtain mutational fingerprints of individual cancer cells. Such data are crucial for the identification of tumour subclones, i. e. genetically diverse cell subpopulations, which are typically associated with treatment failure and the development of drug resistance. The goals of this in-depth analysis are to tailor cancer treatments to individual patients based on the genetic composition of their tumour and thereby increase treatment efficacy and reduce side effects.
On the computational side of this research, large amounts of low quality data are analysed based on complex models to account for various types of noise. In application, this means that practical runtimes of the involved algorithms often limit the size of problem instances that can be solved in reasonable time. As a result, one typically restricts the problem to simplified models or smaller instances. To avoid such restrictions, it is highly desirable to use HPC techniques to speed up the software developed in computational cancer biology.
In this talk, I will introduce our Bayesian inference scheme for tumour mutation histories based on single-cell sequencing. After a general introduction of the problem, I will focus on the Markov Chain Monte Carlo approach that underlies our inference scheme and present performance gains we achieved through HPC.