Prediction Of Protein-Protein Interaction Residue Pairs Based On Residue Frustration And Machine Learning

The study of protein-protein interaction(PPI)is important in understanding the function of proteins.However,it is still a challenge to investigate the transient protein-protein interactions by experiments.Hence,developing an effective computational method to predict PPIs is of great value in protein research.Statistical potential has a wide range of applications and good performance in protein structure prediction and protein folding prediction,but it has not been applied in PPI prediction.In this paper,we explored the application of frustration,a statistical potential,in PPI prediction.By comparing the energetic contribution of the extra stabilization energy from a given residue pair in the native protein to the statistics of the energies,we got residue pair’s frustration index.By calculating the number of residue pairs with high frustration index,the highly frustrated density,a residuefrustration-based feature,was then obtained to describe the tendency of residues to be involved in PPI.Features related to highly frustrated density,as well as structure-based features,were then used to describe protein residues and combined with Long short-term memory(LSTM)neural network to predict PPI residue pairs.Our model,which considers the statistics-based features,is significantly different from the models based on the physicochemical features of residues and has better performance.Our model correctly predicted 75%dimers when only the top 2‰ residue pairs were selected in each dimer.We found that frustration may supersede the physicochemical features and effectively describe the tendency of residue to be involved in PPI.Some previous studies on frustration only revealed that frustration can explain the mechanism of PPI.In this paper,we found that frustration can also be used to predict PPI and have a good performance.It suggests the great potential of statistical potential such as frustration in PPI prediction.