Recent developments in research on the stability of proteins show that ion-pairs potentially contribute to the thermostability of proteins, according to comparisons of ion-pairs between homologous structures. This study proposes a probabilistic Bayesian statistical method to predict
efficiently the thermostability of proteins based on considering the properties of ion-pairs. The experimental results suggest that the number, types, and bond distance of ion-pairs can be used to predict thermostability of proteins with functions similar to each other with a high accuracy up to 80%. The predictions have high precision of 99%, especially for hyperthermophilic proteins. The experimental results of proteins with different functions also indicate that the number of ion-pairs is related to the thermostability of proteins, and predictions of thermostability can also be made for proteins with different functions.