The goal of this project is to increase the statistical power of current genome-wide association studies (GWAS) for brain disorders and aid in their interpretation. This is done by linking genomic variations to observational features in population imaging studies – both imaging and genetic data – using prior information of spatio-temporal gene expression patterns in the brain.
The Allen Brain Atlas gene expression data is a very rich source on the genetics of the mammalian brain. We have explored the use of t-distributed stochastic neighbourhood embedding (t-SNE) for retrieving structure in this data. The results show that neuroanatomical regions can be consistently separated based on their expression characteristics. The non-linear method t-SNE captures local similarities much better than for instance PCA and MDS and reveals interesting structural patterns (Mahfouz et al. 2015). Also, we have demonstrated a strong relation between chromatic folding structing of chromosomes in the cell nucleus and the spatial co-expression of genes in the mouse cortex as derived from the Allen Brain Atlas (Babei e.a, 2015).