RLBindDeep: A ResNet-LSTM based novel framework for protein-ligand binding affinity prediction.

Area of Science:

• Computational chemistry
• Pharmacology
• Artificial intelligence

Background:

• Accurate prediction of protein-ligand binding affinity is crucial for effective drug discovery.
• Traditional docking methods often lack accuracy due to simplified modeling and interaction considerations.

Purpose of the Study:

• To introduce RLBindDeep, a novel deep learning architecture for enhanced prediction of protein-ligand binding affinities.
• To develop a pose-independent regression model that directly predicts binding affinities from complex structures.

Main Methods:

• RLBindDeep integrates ResNet and LSTM architectures.
• The model extracts ligand physicochemical descriptors, protein features, and interaction energies.
• It operates as a pose-independent regression model, directly predicting affinities from fixed complex structures.

Main Results:

• RLBindDeep achieved a Pearson's R of 0.875, Spearman's ρ of 0.864, and RMSE of 0.993 on the CASF-2016 dataset.
• The model demonstrated superior performance compared to state-of-the-art methods like HAC-Net and AutoDock Vina.
• Extracted features included ligand properties, amino acid composition, and various interaction energies.

Conclusions:

• RLBindDeep significantly improves the accuracy and robustness of binding affinity prediction.
• Deep learning approaches, exemplified by RLBindDeep, have the potential to revolutionize computational drug discovery.
• The model offers a more efficient and targeted strategy for drug development processes.