Output details
15 - General Engineering
King's College London
Discovering genetic associations with high-dimensional neuroimaging phenotypes : A sparse reduced-rank regression approach
This paper introduces a novel statistical approach for carrying out genome-wide association studies using neuroimaging measurements as phenotypes. The methodology enables detection of genetic variants that are predictive of very high-dimensional quantitative traits, and can be used for brain-wide searches. The proposed technique can achieve much higher statistical power compared to traditional univariate methods. This was the first truly multivariate approach for 'neuroimaging genetics' studies, and has been successful in detecting associations in Alzheimer’s disease (AD) [Vounou Neuroimage 2012] and multiple sclerosis [Inkster Neurobiol Aging 2013, Strijbis Mult Scler 2013] including potential new gene targets.