Protein-Ligand Interaction Prediction with Effective Method of Graph Attention Networks and Long Short-Term Memory

Le résumé fourni par la source

Nowadays, the drug discovery field of molecules in protein-ligand interactions developing effective therapeutics for identifying potential drug candidates with spatial features to predict binding affinity. However, the existing Deep Learning (DL) based prediction approaches failed to interact and predict the binding affinity of protein-ligand precisely for drug design with binding affinity. To overcome this problem, a Graph Attention Networks with Long Short-Term Memory (GATN-LSTM) method is proposed with spatiotemporal forecasting for interaction between protein and ligand in binding affinity prediction for drug design. The proposed GATN-LSTM method is employed on DUD-E dataset consists of enhanced the prediction by capturing relationship (interaction) between protein and ligand in molecules for drug discovery using PubChem for prediction. Then, the preprocessed data of protein-ligand interaction is extracted using (CNN-GNN) for better generalization of both local and global features for predicting drug discovery. Finally, GATN-LSTM is employed for interaction of molecules with binding affinity which enhanced the drug design and also improved the prediction results. Experimental results of the proposed GATN-LSTM method attained accuracy (0.9992), F1-score (0.9985) are higher than existing prediction approaches such as Convolutional Neural Network (CNN) and Mutual Information-based Feature Selection (MIFS).

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.