Prediction of pathogenic mutations in mitochondrially encoded human tRNAs

Some mutations in human mitochondrial tRNAs are severely pathogenic. The available computational methods have a poor record of predicting the impact of a tRNA mutation on the phenotype and fitness. Here patterns of evolution at tRNA sites that harbor pathogenic mutations and at sites that harbor phenotypically cryptic polymorphisms were compared. Mutations that are pathogenic to humans occupy more conservative sites, are only rarely fixed in closely related species, and, when located in stem structures, often disrupt Watson-Crick pairing and display signs of compensatory evolution. These observations make it possible to classify ∼90% of all known pathogenic mutations as deleterious together with only ∼30% of polymorphisms. These polymorphisms segregate at frequencies that are more than two times lower than frequencies of polymorphisms classified as benign, indicating that at least ∼30% of known polymorphisms in mitochondrial tRNAs affect fitness negatively.