Logistic matrix factorisation and generative adversarial neural network-based method for predicting drug-target...

Area of Science:

• Computational chemistry
• Bioinformatics
• Drug discovery

Background:

• Accurate drug-target interaction prediction is vital for efficient drug discovery.
• Existing computational models often analyze drug and target features separately, missing crucial interaction information.
• Traditional wet-lab methods for identifying drug-target pairs are resource-intensive.

Purpose of the Study:

• To develop an advanced computational method for predicting drug-target protein association pairs.
• To overcome limitations of existing models by extracting joint interaction features.
• To enhance prediction accuracy and efficiency in drug discovery and repurposing.

Main Methods:

• A three-step approach combining matrix factorization and a generative adversarial network (GAN).
• Matrix factorization extracts joint interaction features from the drug-target interaction matrix.
• GAN is employed for data augmentation, generating synthetic positive samples to balance datasets and improve prediction accuracy.
• A four-layer fully connected neural network is utilized for final classification.

Main Results:

• The proposed method achieved prediction accuracy exceeding 97%, outperforming shallow classifiers and state-of-the-art techniques.
• Experimental results confirmed the effectiveness of the data generation step in improving prediction performance.
• Latent interaction features and data augmentation were shown to be efficient for predicting new drug-target associations and drug repurposing.

Conclusions:

• The developed method offers a highly accurate and efficient approach for drug-target interaction prediction.
• The integration of matrix factorization and GANs provides a robust framework for leveraging interaction data.
• This approach holds significant potential for accelerating drug discovery and repurposing efforts.