Dr. Merridee Wouters1, Dr. Thomas Bradley2, Dr. Philip Smith2, Professor James Brenton2, Professor Susan Ramus1
1University Of New South Wales, Kensington, Australia, 2Cancer Research UK Cambridge Institute, Cambridge, UK
Biography:
After graduating in physics, I worked for a year as a clinical physicist in radiotherapy at the Queensland Radium Institute. During this period I became convinced of the need for molecular therapies for cancer. I completed a PhD in bioinformatics in the School of Physics at UNSW. This work investigated structural interactions in proteins and eventually became a passion to understand proteins as machines. My current work focuses on analysis of data from multiple omics platforms to enable precision medicine approaches to diagnosis and therapeutics.
0000-0002-2121-912X
Abstract:
Introduction: Genomics has underpinned significant improvements in tumour classification and patient survival in phenotypes like breast cancer which can be assessed by microarray. However patient survival for phenotypes dominated by copy number variation, like High Grade Serous Ovarian Cancer (HGSOC), has remained refractory. Shallow whole genome sequencing, where multiple samples are multiplexed in a single sequencing lane, provides a potentially cost-effective means of assessment for precision medicine treatments based on chromosomal instability signatures (CIN).
Methods: Here we describe a bioinformatic workflow for processing these samples at UNSW in a high throughput manner, capable of dealing with over a thousand samples in a single run. Firstly, this involves alignment of the sequences on high performance computers like UNSW’s katana or the National Supercomputer GADI. Postprocessing of bam files to produce segmented absolute copy number objects, while still reasonably computationally intensive, can then be performed on a high-end personal computer when R scripts are deployed using Snakemake. This area is still under development as we move away from normalization using allele frequencies to more reliable methods. Further postprocessing on the reduced dataset to assign chromosomal signatures and do further investigation can then be performed in R. Throughout this process, quality controls are deployed and continue to be developed, including dimension reduction techniques such as umap.
Results and Conclusion: Assignment of CIN using non-negative matrix factorisation of copy number characteristics indicate more structure in HGSOC than the accepted Homologous Recombination Deficient (HRD)/nonHRD dichotomy. Further research should enable better assessment and precision medicine treatments.