Aaron J. Mackey, PhD
Statistical and computational genomics, including next-gen SNP discovery, comparative gene structure prediction, functional RNA-seq and ChIP-seq analyses.
There are two predominant areas of my research that are imminently applicable to continuing cancer research. The first is in the arena of discovery and genotyping of genetic variants from ultra-high throughput sequencing data, searching for variants associated with human diseases including cancer, the so-called “driver” mutations that predispose, trigger and maintain neoplastic biology. The search for such point mutations is hampered by genomic instability in tumor genomes, leading to amplifications in genomic copy number that in turn render SNP resolution by sequencing difficult, due to the mixture of reads from >2 haplotypes. This is further compounded by the difficulty in obtaining “pure” cancer genomes, which are typically contaminated by normal tissue. We have developed statistical methodology that attempts to address this, allowing us to identify (germline and somatic) driver and somatic passenger mutations in tumor samples with mixtures of normal and copy-number variant cancer genomes. We plan to use this approach together with colleagues at the OICR as part of the ICGC (www.icgc.org) effort to characterize the genomic blueprint of 50 different tumor (sub)types across tens of thousands of cancer patients.
Secondly, my lab is embarking on an effort to revolutionize the analysis and mining of large-scale experimental genomic datasets. Current approaches to genomic analysis employ what I call “bottom up” approaches: from the raw data, identify a small number of things (probesets, SNPs, ChIP peaks) that change with statistical significance, then try to put those things together to build hypotheses that reflect both the perturbation (e.g. drug treatment), the observed phenotype (e.g. sensitivity to treatment), and the observed genomic changes. While the bottom-up approach can successfully identify key changes that are responsible for phenotypes, it remains up to the biologist to interpret those changes in a context of known biology, and to develop testable hypotheses. The revolution I propose is to complement, or even replace, bottom-up methods with a top-down philosophy that begins by considering various sources of known biology, cast in a framework of probabilistic network topologies that directly represent mechanistic hypotheses, and uses these networks to incorporate and integrate all available genomic data, both experimental and otherwise. These methods will directly identify novel, testable and actionable mechanistic hypotheses, embedded within known biology, and strongly supported by the data. The genesis of this project was my previous GSK research to identify mechanism of sensitivity/resistance to cytotoxic compounds, and I continue to believe these approaches will most greatly benefit cancer research, particularly in the realm of signal transduction pathways that regulate transcriptional activity. My lab has already begun to collaborate with Dr. Tim Bender using such top-down methods to examine the role of Myb transcription factor in thymocyte development. This is an exciting and challenging new project for which I am actively seeking new collaborators and external funding.