BERT-based Models for Predicting Protein-Protein Interaction Sites
Abstract
In this work, a BERT-based model is proposed for predicting potential PPI sites with the aid of its remarkable contextual understanding feat. Thus, the model was adjusted to a set of positive PPI sites, containing 268 sites in total with the accuracy of 0. 89, precision of 0. 88, recall of 0. IL is 0. 78, accuracy of 0. 85, precision of 0. 83, and F1-score of 0. tested on face images achieving a performance of 87% which is higher than that of conventional models such as CNNs and SVMs. Only hyperparameter tuning and cross-validation have been used to make sure that the model is not overfitting to the data and it can produce reasonable results on unseen data as well. The results of this study also indicate that BERT-based models can be effectively utilized in the bioinformatics field, opening up myriad possibilities for future utilization of computational approaches in the prediction of PPIs.