CL-GNN: Contrastive Learning and Graph Neural Network for Protein-Ligand Binding Affinity Prediction

Area of Science:

• Computational chemistry
• Drug discovery
• Machine learning

Background:

• Accurate prediction of protein-ligand binding affinity is crucial for drug discovery.
• Traditional methods are often expensive and time-consuming.
• There is a need for efficient computational approaches.

Purpose of the Study:

• To introduce a novel self-supervised learning (SSL) framework combining contrastive learning and graph neural networks (CL-GNN) for predicting protein-ligand binding affinities.
• To develop a more efficient computational method for drug discovery.
• To enhance the interpretability of binding affinity prediction models.

Main Methods:

• Utilized a contrastive learning strategy, a form of SSL, on a large dataset of 371,458 unlabeled protein-ligand complexes.
• Employed graph neural networks and molecular graph enhancement techniques for self-supervised learning of protein-ligand interactions.
• Assessed similarity between learned representations using cosine similarity.

Main Results:

• The fine-tuned CL-GNN model achieved competitive performance with high Pearson's correlation coefficients and low root-mean-square errors on benchmark datasets.
• The proposed method outperformed existing machine learning models in predicting binding affinities.
• The approach revealed potential connections between complexes and provided insights into drug mechanisms of action.

Conclusions:

• The CL-GNN framework offers a powerful and efficient approach for predicting protein-ligand binding affinities, accelerating drug development.
• The model's interpretability, enhanced through visualization of key residues and atoms, provides valuable biological insights for drug design.
• This self-supervised learning method demonstrates significant potential for advancing computational drug discovery.