Graph neural pre-training based drug-target affinity prediction.

Area of Science:

• Computational chemistry
• Bioinformatics
• Machine learning in drug discovery

Background:

• Pre-training models excel with large unlabeled datasets but struggle with scarce drug-target interaction data.
• Current methods train drug and target models separately, yielding insufficient features for accurate affinity prediction.

Purpose of the Study:

• To develop an improved graph neural pre-training-based drug-target affinity prediction method (GNPDTA).
• To enhance feature representation by integrating information from both unlabeled and labeled drug-target data.

Main Methods:

• Utilized two pre-training models to extract low-level features from drug atom graphs and target residue graphs.
• Employed 2D convolutional neural networks to generate high-level drug and target representations.
• Integrated these representations into a predictor for drug-target affinity estimation.

Main Results:

• The proposed GNPDTA method demonstrated superior performance compared to existing deep learning approaches.
• The approach effectively leverages both unlabeled and labeled data for enhanced feature extraction.
• Experimental results validate the efficacy and improved accuracy of GNPDTA.

Conclusions:

• GNPDTA offers a robust solution for drug-target affinity prediction by optimizing pre-training strategies.
• The method effectively addresses the challenge of limited labeled data in drug discovery.
• This approach holds significant potential for accelerating the identification of novel drug candidates.