DMHGNN: Double multi-view heterogeneous graph neural network framework for drug-target interaction prediction

Area of Science:

• Computational chemistry and cheminformatics
• Bioinformatics and computational biology
• Artificial intelligence in drug discovery

Background:

• Accurate drug-target interaction (DTI) identification is vital for efficient drug discovery.
• Traditional experimental methods for DTI prediction are time-consuming and costly.
• Existing computational methods often overlook comprehensive drug-protein pair (DPP) information within heterogeneous networks.

Purpose of the Study:

• To propose a novel computational framework, the Double Multi-view Heterogeneous Graph Neural Network (DMHGNN), for improved DTI prediction.
• To address the limitations of conventional methods by comprehensively exploring DPP information.
• To enhance the accuracy and efficiency of DTI prediction in drug discovery pipelines.

Main Methods:

• Development of DMHGNN, integrating two multi-view heterogeneous graph neural networks.
• Utilizing meta-paths and denoising autoencoders for learning from protein- and drug-related heterogeneous networks.
• Employing multi-channel graph convolutional networks (GCNs) for learning from drug-protein pair similarity networks (topology, semantics, collaborative graphs).

Main Results:

• The meta-path-based graph encoder with attention identifies crucial substructures and learns multi-source neighboring features.
• The denoising autoencoder effectively learns drug and protein features from the heterogeneous network.
• DMHGNN demonstrated superior performance compared to state-of-the-art methods in DTI prediction experiments.

Conclusions:

• DMHGNN provides a powerful and effective framework for accurate drug-target interaction prediction.
• The proposed method significantly advances computational approaches in drug discovery.
• This framework offers a promising direction for accelerating the identification of potential drug candidates.