Machine learning methods for somatic genome rearrangement detection

Machine learning methods for somatic genome rearrangement detection

Project details

Structural variants (SVs) are large-scale genomic changes and are an important type of mutation in cancer. SVs can occur through a variety of biological mechanisms leading to insertions, deletions, duplications, inversions, and translocations in the genome. These mutations can cause cancers and affect response to therapy.  

SV calling is a challenging problem (e.g. Cameron et al, Nat Commun 2019, 10:3240) and many methods are available for identifying SVs, but these often need to be hand tuned to account for different biases and noise properties in different datasets, leading to variable performance on different datasets. 

This project will develop machine learning methods that generate the best possible results from whole genome tumour-normal sequencing data for each patient. We have demonstrated that this idea works well with single pure samples, such as a germline or cell line. To apply this to cancer will involve handling purity, ploidy and subclonality. Solutions to the above problems will be developed and evaluated, with opportunities for application to whole genome sequencing data from the Stafford Fox Rare Cancer project and other cohorts. 

About our research group

The Papenfuss lab undertakes computational biology and bioinformatics research in the Bioinformatics division at WEHI.

We develop and apply mathematical, statistical and computational approaches to make sense of different types of omics data from cancer and other diseases in order to drive discoveries. A key focus of the lab is using computational approaches to understanding how cancers are initiated and evolve as they progress and adapt to new environmental niches and in response to therapy.  


Email supervisors



Professor Tony Papenfuss

Tony Papenfuss
Laboratory Head; Leader, Computational Biology Theme
Dr Justin Bedo
Bioinformatics division
Daniel Cameron
Bioinformatics division

Project Type: