Convergent coexpression of autism associated genes suggests some novel risk genes may not be detectable in large-scale genetic studies

Autism spectrum disorder (ASD) is a highly heritable neurodevelopmental disorder characterized by deficits in social interactions and communication. Protein function altering variants in many genes have been shown to contribute to ASD risk; however, understanding the biological convergence across so many genes has been difficult and genetic studies depending on presence of deleterious variation may be limited in implicating highly intolerant genes with shorter coding sequences. Here, we demonstrate that coexpression patterns from human post-mortem brain samples (N = 993) are significantly correlated with the transcriptional consequences of CRISPR perturbations (gene editing, interference and activation) in human neurons (N = 17). Across 71 ASD risk genes, there is significant tissue-specific transcriptional convergence that implicates synaptic pathways. Tissue specific convergence of risk genes is a generalizable phenomenon, shown additionally in schizophrenia (brain) and atrial fibrillation (heart). The degree of this convergence in ASD is significantly correlated with the level of association to ASD from sequencing studies (rho = -0.32, P = 3.03 ×10-65) as well as differential expression in post-mortem ASD brains (rho = -0.23, P = 2.39×10-43). After removing all genes statistically associated with ASD, the remaining positively convergent genes showed intolerance to functional mutations, had shorter coding lengths than the ASD genes and were enriched for genes with clinical reports of potential pathogenic contribution to ASD. These results indicate that leveraging convergent coexpression can identify potentially novel risk genes that are unlikely to be discovered by sequencing studies. Overall, this work provides a simple approach to functionally proxy CRISPR perturbation, demonstrates significant context-specific transcriptional convergence among known risk genes of multiple diseases, and proposes novel ASD risk gene candidates. Overall design: Bulk RNAseq based transcriptomic analyses of human iPSC-derived induced neuron cell lines with CRISPR/Cas9-edited SCN2A heterozygous deletion and wild-type SCN2A.