CSConv2d: A 2-D Structural Convolution Neural Network with a Channel and Spatial Attention Mechanism for...

Area of Science:

• Computational chemistry
• Bioinformatics
• Machine learning

Background:

• Accurate prediction of protein-ligand binding affinity is crucial for efficient drug discovery and repositioning.
• Traditional chemical methods for determining binding affinity are often resource-intensive and time-consuming.
• Developing computational models is essential to accelerate the identification of potential drug candidates.

Purpose of the Study:

• To introduce a novel deep learning method, CSConv2d, for predicting protein-ligand interactions.
• To enhance the DEEPScreen model by incorporating a channel and spatial attention mechanism (CS).
• To evaluate the performance of CSConv2d against existing state-of-the-art drug-target interaction prediction methods.

Main Methods:

• Utilized a deep learning architecture (CSConv2d) with 2-D structural representations of compounds as input.
• Integrated a channel and spatial attention mechanism (CS) into the feature abstraction layers.
• Conducted experiments on the ChEMBLv23 dataset to assess predictive accuracy.
• Validated the model's robustness using docking results for a specific protein-ligand pair (PDB ID: 5ceo, Chemical ID: 50D) and kinase inhibitors.

Main Results:

• CSConv2d demonstrated superior performance in predicting protein-ligand binding affinity compared to the original DEEPScreen model.
• The proposed method outperformed several state-of-the-art drug-target interaction (DTI) prediction models, including DeepConv-DTI, CPI-Prediction, and DeepGS.
• Experimental validation confirmed the robustness and practical applicability of the CSConv2d model.

Conclusions:

• CSConv2d represents a significant advancement in computational approaches for predicting protein-ligand binding affinity.
• The integration of channel and spatial attention mechanisms effectively improves prediction accuracy.
• This novel deep learning method offers a promising tool for accelerating drug discovery and development processes.