MitImpact 4: a genome-wide resource for the interpretation of coding and non-coding mitochondrial DNA variants

Abstract

Mitochondrial DNA variants are a frequent cause of inherited disease, yet their computational interpretation has focused almost exclusively on the 13 protein-coding genes. Variants in tRNA and rRNA genes, which carry a disproportionate share of the known pathogenic alleles, have remained a diagnostic blind spot, because their effect depends on RNA structure rather than on protein sequence and few dedicated predictors exist. MitImpact 4 removes this asymmetry. The database now covers all 49 704 single-nucleotide variants that can occur in the mitochondrial genome and reports their functional consequence, phylogenetic conservation, local mutational constraint, and population frequency in gnomAD, HelixMTdb, and ToMMo. Missense variants are annotated with APOGEE 2 and with newly added predictors, including AlphaMissense, ESM1b, VARITY, and LIST-S2, while tRNA and rRNA variants receive predictions from nAPOGEE, our RNA-specific model, alongside MitoTIP and PON-mt-tRNA. For both classes, scores are accompanied by posterior probabilities of pathogenicity that can be used within the ACMG/AMP framework for mitochondrial DNA. Clinical annotation from ClinVar and MITOMAP, including ClinGen expert-panel classifications, is integrated in a redesigned interface that also annotates user-submitted VCF files. MitImpact 4 is freely available at https://mitimpact.mcb2lab.org.