Comparison of three bioinformatics tools in the detection of ASD candidate variants from whole exome sequencing data

Abstract

Background Autism spectrum disorder (ASD) is a heterogenous multifactorial neurodevelopmental condition with a significant genetic susceptibility component. Thus, identifying genetic variations associated with ASD is a complex task. Whole exome sequencing (WES) is an effective approach for detecting extremely rare protein-coding single nucleotide variants (SNVs). However, interpreting the functional and clinical consequences of these variants requires integrating multifaceted genomic information. Methods We compared the effectiveness of three bioinformatics tools in detecting ASD candidate SNVs from WES data of 250 ASD family trios registered in the National Autism Database of Israel. We studied only rare (less than 1% population frequency), proband specific SNVs. The pathogenicity of SNVs, according to the American College of Medical Genetics (ACMG) guidelines, was evaluated by the InterVar and TAPES tools. In addition, likely gene disrupting (LGD) SNVs were detected based on an inhouse bioinformatics tool, designated PsiVariant, that integrates results from seven insilico prediction tools. We compared the effectiveness of these three approaches and their combinations in detecting SNVs in high confidence ASD genes. Results Overall, 605 SNVs in 499 genes distributed in 193 probands were detected by these tools. The overlap between the tools was 64.1%, 17.0%, and 21.6% for InterVar and TAPES, InterVar and PsiVariant, and TAPES and PsiVariant, respectively. The intersection between InterVar and PsiVariant (I and P) was the most effective approach in detecting ASD genes (OR = 5.38, 95% C.I. = 3.25 to 8.53). This combination detected 102 SNVs in 99 genes among 80 probands (approximate 36% diagnostic yield). Conclusions Our results suggest that integration of different variant interpretation approaches in detecting ASD candidate SNVs from WES data is superior to each approach alone. Inclusion of additional criteria could further improve the detection of ASD candidate variants.