High-resolution prediction of mouse brain connectivity using gene expression patterns

The brain is a multi-level system in which the high-level functions are generated by low-level genetic mechanisms. Thus, elucidating the relationship among multiple brain levels via correlative and predictive analytics is an important area in brain research. Currently, studies in multiple species have indicated that the spatiotemporal gene expression patterns are predictive of brain wiring. Specifically, results on the worm Caenorhabditis elegans have shown that the prediction of neuronal connectivity using gene expression signatures yielded statistically significant results. Recent studies on the mammalian brain produced similar results at the coarse regional level. In this study, we provide the first high-resolution, large-scale integrative analysis of the transcriptome and connectome in a single mammalian brain at a fine voxel level. By using the Allen Brain Atlas data, we predict voxel-level brain connectivity based on the gene expressions in the adult mouse brain. We employ regularized models to show that gene expression is predictive of connectivity at the voxel-level with an accuracy of 93%. We also identify a set of genes playing the most important role in connectivity prediction. We use only this small number of genes to predict the brain wiring with an accuracy over 80%. We discover that these important genes are enriched in neurons as compared to glia, and they perform connectivity-related functions. We perform several interesting correlative studies to further elucidate the transcriptome-connectome relationship.