A probabilistic method to identify compensatory substitutions for pathogenic mutations

Complex systems of interactions govern, the structure and function of biomolecules. Mutations that substantially disrupt these interactions are deleterious and should not persist under selection. Yet, several instances have been reported where a variant confirmed as pathogenic in one species is fixed in the orthologs of other species. Here we introduce a novel method for detecting compensatory substitutions for these so-called compensated pathogenic deviations (CPDs), incorporating knowledge of pathogenic variants into a probabilistic method for detecting correlated evolution. The success of this approach is demonstrated for 26 of 31 CPDs observed in mitochondrial transfer RNAs and for one in beta hemoglobin. The detection of multiple compensatory sites is demonstrated for two of these CPDs. The methodology is applicable to comparative sequence data for biomolecules expressed in any alphabet, real or abstract. It provides a widely applicable approach to the prediction of compensatory substitutions for CPDs, avoiding any reliance on rigid non-probabilistic criteria or structural data. The detection of compensatory substitutions that facilitate the substitution of otherwise pathogenic variants offers valuable insight into the molecular constraints imposed on adaptive evolution.