Prediction of transient and permanent protein interactions using AI methods

Protein-protein interactions (PPIs) can be classified as permanent or transient based on their stability lifetime. Understanding the precise details of such protein will pave way for discovery inhibitors and understanding nature function PPIs. In present work, 43 relevant physicochemical, geometrical structural features were calculated a curated dataset from literature, comprising 402 protein-protein complexes categories, 5 different Supervised Machine Learning models developed with Scikit-learn to predict PPI. Additionally, deep learning method Artificial Neural Network was also performed using Tensor Flow Keras. Predicted achieved accuracy ranging 76.54% 82.71% k-NN has highest accuracy. Detailed analysis these methods revealed that Interface areas Percent interface accessible area, area Total parameters defining shape PPI Planarity, Eccentricity Circularity are most discriminating factors between two categories. The could serve an effective tool understand mechanism association interactions, which supplement costly time-consuming experimental techniques.