MAGNETIC: Multilayer Attention and Graph Neural Network Diffusion for Effective Drug-Target Interaction Prediction

Area of Science:

• Bioinformatics
• Computational Drug Discovery
• Network Science

Background:

• Drug-target interaction (DTI) prediction is crucial for accelerating drug discovery and reducing wet lab validation costs.
• Conventional methods often struggle with the complexity and sparsity of DTI data.
• Graph-based approaches offer a promising avenue for modeling intricate biological networks.

Purpose of the Study:

• To introduce MAG NETIC (Multilayer Attention and Graph Neural nETwork dIffusion for effeCtive Drug-Target Interaction prediction), a novel framework for enhanced DTI prediction.
• To leverage multilayer attributed heterogeneous graphs and diffusion-based graph neural networks (GNNs) for improved DTI prediction.
• To validate the efficacy of MAG NETIC against existing state-of-the-art methods.

Main Methods:

• Representing drugs and targets as a multiplex network incorporating meta-path-derived similarity relations.
• Employing separate Adaptive Graph Diffusion Networks (AGDNs) per layer to learn layer-specific representations.
• Utilizing a graph auto-encoder framework with a reconstruction objective tailored for sparse and imbalanced DTI data.

Main Results:

• MAG NETIC consistently outperformed baseline methods in predicting drug-target interactions, as measured by AUROC and AUPRC.
• Significant improvements were observed in AUPRC, highlighting the model's effectiveness in identifying and ranking true interactions within imbalanced datasets.
• Real-world validation demonstrated the practical utility of MAG NETIC for discovering novel DTIs.

Conclusions:

• Modeling DTI as a multilayer attributed heterogeneous graph with layer-specific GNNs significantly enhances predictive performance.
• MAG NETIC offers a robust and effective approach for accelerating drug discovery by improving DTI prediction accuracy and ranking.
• The framework shows practical utility in identifying previously undiscovered drug-target interactions, aiding in novel drug development.