|
Abstract:
|
With the availability of large-scale brain imaging and genetics data, brain imaging genetics, which studies the relationship between genetic variations and brain imaging phenotypes, is becoming an emerging and rapidly growing research field. Among different imaging quantitative traits, brain connectivity/network is the one plays an essential role to shape physiological and pathological behaviors as well as brain functional or structural composition. However, there are still limited works investigate the impact of genetic variants on brain connectivity and the existing ones typically broke down the contingency matrix via vectorization, ignoring the natural network structure. In this work, we developed a biologically interpretable network response regression under a proposed Bayesian hierarchical shrinkage prior. Our model incorporated the LD information within SNPs as well as the population brain network to better understand which and how genotypes impact the individual level brain connectivity. We also develop an efficient EM algorithm which scales up our method to analyze the whole brain network. We applied our method to Human Connectome Project (HCP).
|
