Automated sequence variant classification tool for DNA diagnostics

Abstract

Advancements in DNA sequencing rapidly improved our understanding of the genome in recent years. Today, these advances are revealing thousands of genetic variants that are still waiting to be deciphered. Establishing the association between genetic variation and diseases enables us to better appreciate the biology of diseases and to develop effective therapeutical solutions. To this end, clinical organizations such as American College of Medical Genetics and Genomics (ACMG) developed standards and guidelines to interpret sequence variants. Clinical Genomics Resource (ClinGen) provided further specifications to the guidelines to improve the interpretations. However, implementation of the guidelines takes considerable time and requires substantial expertise in clinical genetics. Available computational tools to automate the process (i) do not comprehensively describe how their frameworks function, (ii) fail to completely follow the latest specifications and (iii) lack high consistency with variant classifications manually performed by experts. Here, this work presents automated ACMGbased variant classifier (AAVC), which computationally interprets sequence variants based on the ACMG Guidelines and the ClinGen Specifications by aggregating information from large public databases and in silico prediction tools including BayesDel, ClinVar, Ensembl, gnomAD, PhyloP, RepeatMasker, SpliceAI and UniProt. The tool demonstrates a high concordance (99.67%) with FDA-approved variant classification database, reveals more than two hundred novel variants in clinically actionable genes in the Turkish Variome and reclassifies at least 57,000 inconclusive variants in ClinVar as pathogenic or likely pathogenic. The work provides a comprehensive framework to enable rapid and accurate interpretation of sequence variants by the ACMG Standards.