Autism as a neurodevelopmental disorder causes significant reduction in reproductive rate among affected individuals. As a result, genetic risk variants with large effect on autism are under strong selection, which can be quantified as selection coefficient. Yufeng Shen and colleagues have recently published methods to estimate selection coefficients for all missense variants and protein truncating variants.

Recent genomic studies of autism have discovered over 200 high-confidence autism risk genes, mostly based on de novo coding variants and with pleiotropic effect in other neurodevelopmental disorders. The Shen lab has shown that most of the population attributable risk from de novo coding variants is in known autism genes, whereas PAR from rare inherited variants is mostly in unknown risk genes. This indicates that the genes that contribute to autism risk predominantly through rare inherited variants may point to different molecular and neurodevelopmental pathways. Identification of these risk genes and variants in autism individuals and cohorts is important for better understanding of the severity, clinical trajectories of various conditions, and optimized interventions.

Previous studies have largely dependent on de novo variants. Shen and colleagues argue that selection coefficient is a more direct metric related to genetic effect. Their hypothesis is that selection coefficient can be a universal metric that can be applied to both de novo and rare inherited variants in genetic analysis of early conditions like autism. In this study, they aim to elucidate genetic architecture of autism through the lens of natural selection. Additionally, we will use selection coefficient to improve identification of genetic causes in autistic individuals, especially for the ones whose parental genomic data is not available.