Integration of estimated regional gene expression with neuroimaging and clinical phenotypes at biobank scale

Abstract

Recent studies have revealed mappings between brain-wide gene expression profiles and a range of brain structure and activity phenotypes. Most of these studies have used group-averaged expression data. While informative, this approach cannot capture the high degree of individual variation in brain phenotypes. The investigation of this variation to date has been limited by availability of gene expression and neuroimaging data for many individuals. Here, we addressed this limitation by adopting PrediXcan, an established framework for inferring genetically regulated gene expression using localized genetic variants. This framework has allowed us to study the otherwise inaccessible expression profiles for brain regions across many individuals. We used this approach to identify genes whose inferred expression across brain regions of interest correlated with a range of phenotypes. Our analyses bridge a gap in human neuroimaging by enabling the study of associations between individual-level gene expression and brain structure and activity. Ultimately, this approach can help reveal mechanistic pathways from gene expression to healthy or diseased brain function.