Principal and Independent Genomic Components of Brain Structure and Function: Reproducibility and Relevance

COEX 

Individual differences in brain structure and function, as measured using MRI, are heritable. Genome-wide association studies (GWAS) are conducted to gain an understanding of the molecular mechanisms driving this inheritance of brain variation. However, this mechanistic translation is challenging due to the high polygenicity and pleiotropy. The typical big data structure we have with neuroimaging genomics is a very large number of traits (across brain regions and MRI modalities) associated with millions of genomic variants, in a way that shows intricate cross-trait genetic correlation patterns. Taking advantage of this high-dimensional data structure, we developed genomic independent and principal component analysis (ICA and PCA) to decompose thousands of GWASs of neuroimaging traits into more interpretable genome-wide components. Our results show that 10 components explain ~40% of the total genetic associations across >2,000 neuroimaging traits, and improve the reproducibility of the genetic signal. Further, we show that several of these components index loci with shared biological functions and/or behavioural associations, and that they are mainly driven by distinct genetic underpinnings across tissues and MRI modalities. Currently, we are investigating how the different genomic components map to genetic risk for brain disorders. Taken together, these results encourage further applications of genomic ICA and PCA to high-dimensional GWAS data to improve interpretability, signal-to-noise ratio, and potentially individual stratification of healthy and clinical populations.