Despite the use of genome-wide sequencing, families who present to genetics clinics seeking to understand the cause of the autism or intellectual disability (ID) in their family member only receive a genetic diagnosis in approximately 10 to 15 percent of cases.

The John Greally group is leveraging its experience in epigenetics to study the influence of functional non-coding variants (FNCVs) in neurodevelopment conditions. Their preliminary study of 20,366 genomes has revealed significant accumulation of de novo variants (DNVs) in regulatory regions of glutamatergic neurons and a burden effect in individuals with autism accompanied by ID, in whom an increasing number of DNVs in regulatory regions of glutamatergic neurons was associated with decreasing IQ.

In this proposal, a novel strategy is described to increase the cohort size to 434,429 individuals, in each of whom many more candidate FNCVs become available for analysis, representing what would be the largest study to date of FNCVs in a human disease or trait.

Updated annotations of candidate FNCVs will include the use of the bias-factorized neural network ChromBPNet and the deep learning model AlphaGenome. For studies focused on the individual in the cohort, the group will assemble their annotations and perform a Random Forest machine learning classifier to model those with autism and/or ID, and then use counterfactual and adversarial analysis to examine the impact of specific features on prediction.

The group’s foundation study indicates that a substantial proportion of the unsolved individuals with autism and/or ID harbour FNCVs in regulatory regions of glutamatergic neurons. The proposed expanded study would be substantially more powerful and would deliver (a) insights into the pathobiology and genomic architecture of autism and/or ID, and (b) the basis for a genetic diagnostic testing approach to understand FNCV effects.