MGACL: Prediction Drug-Protein Interaction Based on Meta-Graph Association-Aware Contrastive Learning

Area of Science:

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

Background:

• Accurate drug-target interaction (DTI) identification is essential for efficient drug discovery.
• Graph neural networks (GNNs) show promise for DTI prediction but suffer from bias and negative transfer.
• Existing methods struggle to effectively leverage heterogeneous auxiliary information for personalized DTI prediction.

Purpose of the Study:

• To develop an adaptive network that mitigates bias and negative transfer in GNN-based DTI prediction.
• To effectively transfer personalized heterogeneous auxiliary information for improved DTI identification.
• To enhance the accuracy of drug-target interaction prediction for novel drug development.

Main Methods:

• Introduction of the Meta Graph Association-Aware Contrastive Learning (MGACL) network.
• Implementation of an adaptive personalized meta-knowledge transfer mechanism for heterogeneous auxiliary information.
• Proposal of a novel DTI association-aware contrastive learning strategy to align drug representations and prevent negative transfer.

Main Results:

• The proposed MGACL network effectively transfers personalized heterogeneous auxiliary information.
• The contrastive learning strategy successfully prevents negative transfer by aligning representations.
• MGACL achieved an approximate 3% improvement in DTI prediction performance, as indicated by AUC and AUPRC metrics, outperforming existing methods.

Conclusions:

• MGACL offers a robust framework for enhancing GNN-based DTI prediction accuracy.
• The adaptive meta-knowledge transfer and contrastive learning approach effectively addresses bias and negative transfer.
• This advancement facilitates more precise identification of drug targets, accelerating new drug development.