Discrimination and Prediction of Protein-Protein Binding Affinity Using Deep Learning Approach

Protein-protein interactions (PPIs) mediate myriad biological functions. Estimating the binding affinity for PPIs help us to understand the underlying molecular recognition mechanism. In this work, we utilized deep learning approach to discriminate protein-protein complexes based on their binding affinity. We setup a database of 464 protein-protein complexes along with their experimental binding affinities and developed a deep learning based binary classification model, which showed an accuracy of 81.75% using 5-fold cross-validation. Furthermore, we refined the method for predicting the binding affinity of protein-protein complexes using a large set of complexes (PPA-Pred2). It could predict the binding affinity of the training set and blind test set with a mean absolute error of 1.24 kcal/mol and 1.31 kcal/mol, respectively. We suggest that our methods could serve as efficient tools to study PPIs and provide crucial insights about the underlying mechanism for the molecular recognition process.