Databases' Variant Classifications Have Improved Over Time

The accuracy of variant classifications in databases has gotten with time, according to a new analysis appearing in Genome Medicine. Researchers from the University of California, Berkeley, examined changes in variant misclassification within archived versions of ClinVar and the Human Gene Mutation Database (HGMD) over six years, focusing in particular on variants associated with inborn errors of metabolism. From within the 1,000 Genomes Project, the researchers identified individuals with genotypes that the databases said were pathogenic. As these conditions are rare and develop among infants, the researchers note these classifications within the 1,000 Genomes Project — a cohort of adults — likely represented misclassifications in the databases. These misclassifications have decreased within both ClinVar and HGMD over time, they write, adding that ClinVar, where reclassification are more frequent, has a lower false-positive rate. "We anticipate that this work will be a valuable benchmark of the progress that has been made in variant interpretation, of interest to the individuals who maintain these databases, the clinical laboratories and researchers who use these databases regularly, and the computational researchers who use these databases for training and testing methods," the researchers write.