Genome-Scale Transcriptional Regulatory Network Models of Psychiatric Disorders

Abstract

Non-coding genetic variants with predicted regulatory functions contribute strongly to risk for psychiatric disorders, and psychiatric disorders are associated with changes in the expression of hundreds of genes in the brain. Understanding brain gene regulation may therefore provide insights into the mechanisms of psychiatric disorders, yet mechanistically and quantitatively accurate models of human transcriptional regulatory networks remain elusive. I will describe TRENA, a new strategy to predict tissue-specific binding sites and functional targets genes for hundreds of transcription factors (TFs) by integrating large epigenomic and transcriptomics datasets. We used TRENA to model TF-target gene interactions in the human brain, combining brain-specific ENCODE data with 2,756 gene expression profiles from the Allen Brain Atlas. This model made quantitatively accurate predictions (r2 > 0.5) about the expression patterns of 11,093 genes based on the binding sites and expression patterns of 778 TFs. Integrating this TRN model with post-mortem gene expression profiles from the prefrontal cortex of psychiatric cases and controls led to the identification of 66 master regulator TFs that are predicted to drive expression changes in both bipolar disorder and schizophrenia. Our model also predicted regulatory functions for SNPs associated with genetic risk for bipolar disorder and schizophrenia. We experimentally validated our model’s prediction that a risk-associated SNP alters the activity of the VRK2 promoter and occupancy of a master regulator TF, POU3F2.