Drug-target binding affinity prediction based on power graph and word2vec.

Area of Science:

• Computational chemistry
• Bioinformatics
• Machine learning in drug discovery

Background:

• Drug and protein targets interact to influence physiological and metabolic functions.
• Accurate prediction of drug-target interactions is critical for efficient drug development.
• Machine learning approaches are increasingly utilized to accelerate drug discovery and reduce costs.

Purpose of the Study:

• To propose a novel drug target affinity prediction model named WPGraphDTA.
• To leverage power graph and Word2vec techniques for enhanced prediction accuracy.
• To reduce the time and expenses associated with drug retargeting and new drug development.

Main Methods:

• Utilizing a graph neural network to extract drug molecular features within a power graph module.
• Employing the Word2vec method to obtain protein features.
• Integrating drug and protein features followed by processing through three fully connected layers for affinity prediction.

Main Results:

• The proposed WPGraphDTA model demonstrates strong predictive performance on benchmark datasets.
• Regression-based drug target affinity prediction proves more representative of binding ability.
• The model achieved good prediction performance on the Davis and Kiba datasets.

Conclusions:

• WPGraphDTA effectively predicts drug target affinity by combining graph neural networks and Word2vec.
• The model's performance indicates its potential to streamline the drug development pipeline.
• Accurate affinity prediction using WPGraphDTA can significantly cut down drug development timelines and costs.