Predicting functional neuroanatomical maps from fusing brain networks with genetic information

Abstract

A central aim, from basic neuroscience to psychiatry, is to resolve how genes control brain circuitry and behavior. This is experimentally hard, since most brain functions and behaviors are controlled by multiple genes. In low throughput, one gene at a time, experiments, it is therefore difficult to delineate the neural circuitry through which these sets of genes express their behavioral effects. The increasing amount of publicly available brain and genetic data offers a rich source that could be mined to address this problem computationally. However, most computational approaches are not tailored to reflect functional synergies in brain circuitry accumulating within sets of genes. Here, we developed an algorithm that fuses gene expression and connectivity data with functional genetic meta data and exploits such cumulative effects to predict neuroanatomical maps for multigenic functions. These maps recapture known functional anatomical annotations from literature and functional MRI data. When applied to meta data from mouse QTLs and human neuropsychiatric databases, our method predicts functional maps underlying behavioral or psychiatric traits. We show that it is possible to predict functional neuroanatomy from mouse and human genetic meta data and provide a discovery tool for high throughput functional exploration of brain anatomy in silico.