PRA-MutPred: Predicting the Effect of Point Mutations in Protein–RNA Complexes Using Structural Features

This is the authors' abstract. We don't add key points for this paper.

Interactions between proteins and RNAs are essential for the proper functioning of cells, and mutations in these molecules may lead to diseases. These protein mutations alter the strength of interactions between the protein and RNA, generally described as binding affinity (Δ G ). Hence, the affinity change upon mutation (ΔΔ G ) is an important parameter for understanding the effect of mutations in protein–RNA complexes. In this work, we developed a machine-learning model to predict ΔΔ G values upon mutations in protein–RNA complexes. We collected experimentally determined ΔΔ G values of 710 mutations in 134 protein–RNA complexes. Diverse sequence and structural features were generated from both wild-type and modeled mutant complexes, which include conservation scores, residue-based, network-based, and interface features. Further, we developed a support vector regressor model with a correlation of 0.75 and a mean absolute error of 0.84 kcal/mol in the jack-knife test. We observed that the performance of the model is dictated by structural features, such as contact potentials, atom contacts in the interface of protein–RNA complexes, and the solvent accessibility of the mutated residue. We also developed a Web server, PRA-MutPred, predicting the protein–RNA binding affinity change upon mutation, which is available in the link https://web.iitm.ac.in/bioinfo2/pramutpred/ .