MGMA-DTI: Drug target interaction prediction using multi-order gated convolution and multi-attention fusion

Area of Science:

• Computational Biology
• Drug Discovery
• Machine Learning

Background:

• Accurate drug-target interaction (DTI) prediction is vital for efficient drug discovery.
• Current deep learning models struggle with global feature extraction and interpretable feature fusion.

Purpose of the Study:

• To develop an advanced deep learning model, MGMA-DTI, for improved DTI prediction.
• To address limitations in global feature learning and feature fusion for enhanced model interpretability.

Main Methods:

• Utilized a graph convolutional neural network for drug feature encoding from SMILES strings.
• Employed multi-order gated convolution for protein sequence global feature extraction.
• Implemented a multi-attention fusion module for effective drug-target interaction feature capture.

Main Results:

• MGMA-DTI significantly outperformed baseline models on BindingDB, BioSNAP, and Human datasets.
• Case studies confirmed the model's utility in providing drug discovery insights.
• The model demonstrated molecular-level interpretability for scientifically meaningful guidance.

Conclusions:

• MGMA-DTI offers a superior approach to DTI prediction by enhancing feature learning and fusion.
• The model's interpretability provides valuable guidance for drug discovery and development.
• This work advances the application of deep learning in pharmaceutical research.