Cancer is not a single disease. Each person's tumour is unique, shaped by the patient's basal germ-line genetics, by the stochastic mutations that cause cancer, and by myriad epigenetic and micro-environmental factors at play inside the human body. Given all these sources of variability, it may be surprising that most patients receive essentially the same treatment, with slight variations.
The underlying goal of our team’s research is to improve survival-rates of cancer patients by predicting for each individual the most effective treatment possible. We do so by using large genomic datasets, such as those generated by microarray experiments or by next-generation genome-sequencing studies (PMID: 26430161). These can be targeted at almost any aspect of the cell, including DNA (methylation, mutations, translocations and copy-number abnormalities), RNA (abundances and splicing events), proteins (abundances or post-translational modifications), and small metabolites (abundances). We develop sophisticated computational models from these datasets, with the aim of predicting clinical behaviour. Our models are then validated independently, and where possible advanced forward as candidate clinical tools.
Key Findings: Biomarkers for Prostate Cancer
Approximately one in seven Canadian men will develop prostate cancer over their lifetime, and the disease is likely to become the most diagnosed cancer in the country over the next decade. Unlike most tumour types, prostate cancer patients are both frequently over-treated (receiving therapies that do not contribute to survival) and under-treated (receiving less therapy than is optimal). Our team focuses on using big-data techniques to better understand prostate cancer biology, and using these insights to develop clinically actionable biomarkers. We have performed the several of the first genomic studies of the disease (PMIDs: 26544944, 26005866) and have used those insights to develop actionable biomarkers integrating measurements of tumour oxygen usage and genomics (PMID: 25456371) and from circulating blood (PMID: 22619380). We are now developing a series of biomarkers using genomic and mitochondrial sequencing, RNA-sequencing, methylation profiling, ChIP-seq, MS-based proteomics and several types of imaging in patient cohorts of several hundred patients. These candidate biomarkers and the underlying biology driven by them will be validated in several patient series of hundreds to thousands of patients to help bring them into clinical practice.
Key Findings: Biological Data Science
To support our work in biomarker development, the team also develops a series of techniques in biological data science. We lead the ICGC-TCGA Dream Somatic Mutation Calling Challenge – a series of crowd-sourced benchmarking projects that are developing the standards for analyzing big biological data (PMIDs: 25314947, 24675517). For example, we have created identified the major sources of error in the identification of point-mutations in DNA sequencing data (PMID: 25984700) in a crowd-sourced challenge supported by Google and Annai bio-systems. Similarly we work on the visualization of big data (http://labs.oicr.on.ca/boutros-lab/software/bpg) and on the quality-assessment of genomic data (PMIDs: 22513995, 23146350, 23169800, 23537167, 25173705).
Students in my lab come from a broad range of research backgrounds, including biochemistry, mathematics, biology, and computer science. Students take on and lead significant projects, gaining a deep understanding of cancer biology, but also of cutting-edge technologies and all within an environment of professional software-development and statistical analysis. Mentorship includes a students-only journal club, regular meetings, and tight integration with clinical and other colleagues to help in networking.