A knowledge distillation-guided equivariant graph neural network for improving protein interaction site prediction performance

Shouzhi Chen, Zhenchao Tang, Linlin You, Calvin Yu-Chian Chen

Abstract: Highlights•The ESM2 is used to represent the features of amino acids in proteins to improve the accuracy and generalization of the model.•The model uses a self-distillation strategy to aid in model training, using information from the model itself to improve its generalization and hierarchical representation.•The PPIS GUI platform was designed based on Python’s PyQt5 library. By designing the GUI, users are able to manipulate the PPIS prediction model more easily.