GIAE-DTI: Predicting Drug-Target Interactions Based on Heterogeneous Network and GIN-Based Graph Autoencoder.

Area of Science:

• Computational chemistry
• Bioinformatics
• Machine learning

Background:

• Accurate drug-target interaction (DTI) prediction is crucial for drug discovery and repurposing.
• Existing computational methods struggle with sparse DTI data, limiting their ability to aggregate neighbor node information and represent isolated nodes.

Purpose of the Study:

• To develop a novel deep learning framework, GIAE-DTI, for enhanced DTI prediction.
• To address the limitations of existing methods in handling sparse DTI data and improve information aggregation.

Main Methods:

• Constructed a heterogeneous network incorporating drug-drug, protein-protein, and weighted K-nearest neighbor processed drug-target interactions.
• Employed a graph autoencoder with a graph isomorphism network for feature extraction and a dual decoder for self-supervised learning.
• Utilized a deep neural network for final DTI prediction based on learned latent representations.

Main Results:

• GIAE-DTI achieved AUC of 0.9533 and AUPR of 0.9619 on a benchmark dataset, surpassing current state-of-the-art methods.
• Demonstrated practical applicability through case studies involving 5-hydroxytryptamine receptor targets and drugs for mental diseases.

Conclusions:

• GIAE-DTI offers a robust and effective approach for DTI prediction, particularly in scenarios with sparse data.
• The framework shows significant potential for accelerating drug discovery and repurposing efforts.