It is possible to extrapolate pathogenicity in mutations in the same position in homologous proteins

Abstract

Abstract When we compare an individual’s genome with the reference, several mutations are encountered. Most of these mutations are neutral, but some others can lead to pathogenic consequences. Given the rapid increase in the amount of generated sequencing data, there is an urgent need to accurately determine whether genetic variants detected in patients are disease causing or not. While numerous computational predictive tools exist, their ability to make accurate predictions is still limited. In this study, we focus on missense variants, those that modify the coding amino acid, and our aim is to determine if the pathogenicity of these variants can be extrapolated to homologous variants, i.e., variants affecting the same position in homologous proteins and exhibiting the same or similar amino acid change. With this purpose, we extracted homologous variants in a dataset composed of all reported disease-causing (ClinVar) and neutral (gnomAD) human missense variants from proteins with autosomal dominant (AD) inheritance. We collected 63,192 pairs of homologous variants from which 60,822 were disease-causing, 1,799 were neutral and 571 of them disagreed in pathogenicity annotation, achieving an error rate of 0.9%. Thus, our data supports that pathogenicity can be extrapolated, with a high accuracy and reliability, between homologous variants. This approach expands the number of variants for which pathogenicity can be annotated with a high precision.