Causal enhanced drug-target interaction prediction based on graph generation and multi-source information fusion

Area of Science:

• Biomedical Informatics
• Computational Biology
• Drug Discovery

Background:

• Drug-target interaction (DTI) prediction is crucial for developing effective therapies.
• Existing DTI methods struggle with interpretability and require manual feature engineering.
• Graph generation offers flexible information fusion for DTI prediction.

Purpose of the Study:

• To develop a novel causal enhanced method for drug-target interaction (CE-DTI) prediction.
• To improve the accuracy and interpretability of DTI prediction models.
• To leverage graph generation and multi-source information fusion for enhanced DTI prediction.

Main Methods:

• Representing drugs and targets via fused multi-source information through automatic graph generation.
• Constructing a drug-target pairs network for node classification.
• Separating causal and non-causal variable nodes and applying causal invariance for contrastive learning.

Main Results:

• The proposed CE-DTI method achieved superior performance compared to benchmark methods across multiple datasets.
• The causal enhancement strategy effectively identified potential causal effects between drug-target pairs.
• The method demonstrated success in discovering new potential drug targets.

Conclusions:

• CE-DTI provides an effective and interpretable approach for predicting drug-target interactions.
• The causal enhancement strategy aids in uncovering novel therapeutic targets.
• This method facilitates targeted therapy development with improved efficacy and reduced side effects.