Whole-genome deep-learning analysis identifies contribution of noncoding mutations to autism risk

Abstract

We address the challenge of detecting the contribution of noncoding mutations to disease with a deep-learning-based framework that predicts specific regulatory effects and the deleterious impact of genetic variants. Applying this framework to 1,790 Autism Spectrum Disorder (ASD) simplex families reveals disease causality of noncoding mutations: ASD probands harbor both transcriptional and post-transcriptional regulation-disrupting de novo mutations of significantly higher functional impact than unaffected siblings. Further analysis suggests involvement of noncoding mutations in synaptic transmission and neuronal development, and taken together with prior studies reveal a convergent genetic landscape of coding and noncoding mutations in ASD. We demonstrate that sequences carrying prioritized proband mutations possess allele-specific regulatory activity, and highlight a link between noncoding mutations and IQ heterogeneity in ASD probands. Our predictive genomics framework illuminates the role of noncoding mutations in ASD, prioritizes high impact mutations for further study, and is broadly applicable to complex human diseases.

Evaluating the informativeness of deep learning annotations for human complex diseases

“…Our work has several limitations, representing important directions for future research. First, our analyses of deep learning annotations using S-LDSC are inherently focused on common variants, but deep learning models have also shown promise in prioritizing rare pathogenic variants 15,38,39 . The value of deep learning models for prioritizing rare pathogenic variants has been questioned in a recent analysis focusing on Human Gene Mutation Database (HGMD) variants 40 , meriting further investigation.…”

Cross-species regulatory sequence activity prediction

“…In these data, the offspring have an average of 67 de novo mutations, which have a slight enrichment in promoters (29). Recent work demonstrated that variant effect predictions further differentiate autism cases from their unaffected sibling controls (30). We hypothesized Figure 5: Human de novo variant predictions for mouse data enrich for autism cases versus controls.…”

Genome-wide patterns ofde novotandem repeat mutations and their contribution to autism spectrum disorders

“…We present a novel framework for identification of de novo TR mutations and apply our method to WGS of more than 7,000 SSC samples. Our analysis reveals on average 54 autosomal TR mutations per child, which is similar in magnitude to the total de novo burden of all point mutations 5,10,48 . Notably, we likely underestimate actual mutation rates due to the stringent filtering applied to candidate mutations, which limited the total fraction of genomic TRs included in our analysis.…”

“…This excess is consistently observed across SSC phases and remains after controlling for paternal age. Based on this excess we estimate that autosomal TR mutations contribute to approximately 1.6% of simplex idiopathic ASD probands, comparable in magnitude for non-coding point mutations 10 . This number will likely increase when sex chromosomes and more difficult to genotype TRs are included.…”

“…Through massive next-generation sequencing efforts, de novo point mutations in over 102 genes are now robustly associated with ASD 9 . Yet, mutations in these genes still explain a minority of cases 10 , and recent studies suggest many specific mutations leading to ASD remain to be found 11,12 .…”