An integrative scoring approach for prioritization of rare autism spectrum disorder candidate variants from whole exome sequencing data

Discerning clinically relevant autism spectrum disorder (ASD) candidate variants from whole-exome sequencing (WES) data is complex, time-consuming, and labor-intensive. To this end, we developed AutScore, an integrative prioritization algorithm of ASD candidate variants from WES data and assessed its performance to detect clinically relevant variants. We studied WES data from 581 ASD probands, and their parents registered in the Azrieli National Center database for Autism and Neurodevelopment Research. We focused on rare allele frequency (< 1%) and high-quality proband-specific variants affecting genes associated with ASD or other neurodevelopmental disorders (NDDs). We developed AutScore and AutScore.r and assigned each variant based on their pathogenicity, clinical relevance, gene-disease association, and inheritance patterns. Finally, we compared the performance of both AutScore versions with the rating of clinical experts and the NDD variant prioritization algorithm, AutoCaSc. Overall, 1161 rare variants distributed in 687 genes in 441 ASD probands were evaluated by AutScore with scores ranging from - 4 to 25, with a mean ± SD of 5.89 ± 4.18. AutScore.r cut-off of ≥ 0.335 performs better than AutoCaSc and AutScore in detecting clinically relevant ASD variants, with a detection accuracy rate of 85% and an overall diagnostic yield of 10.3%. Five variants with AutScore.r of ≥ 0.335 were distributed in five novel ASD candidate genes. AutScore.r is an effective automated ranking system for ASD candidate variants that could be implemented in ASD clinical genetics pipelines.