Presentation: Towards Modern Machine Learning for Genome-wide Association Studies
Speaker: Lloyd Elliott, PhD, Department of Statistics and Actuarial Sciences, Simon Fraser University
Abstract: In collaboration with the University of Oxford Department of Statistics and the Wellcome Centre for Integrative Neuroimaging, a genome-wide association study was conducted on 3,144 phenotypes derived from brain magnetic resonance imaging (MRI) in 8,428 participants of the UK Biobank consortium. Classical analyses of genetic effects can provide heritability, genetic correlation and enrichment analyses. We discuss the results from classical analyses, and provide directions for genome-wide association studies on neuroimaging phenotypes such as brain age and voxelwise approaches. We also discuss how to bring genome-wide association studies into the predictive frameworks used by modern machine learning methods such as Bayesian statistics and Deep Learning, and outline new frontiers in digital medicine.