Semantic-Enhanced Graph Contrastive Learning With Adaptive Denoising for Drug Repositioning

Area of Science:

• Computational chemistry
• Bioinformatics
• Drug discovery

Background:

• Traditional drug development is resource-intensive.
• Drug repositioning offers an efficient alternative but faces challenges.
• Existing methods struggle with sparse datasets and data noise.

Purpose of the Study:

• To develop an advanced drug repositioning model.
• To address data sparsity and noise in drug repositioning datasets.
• To enhance the accuracy and robustness of drug repositioning predictions.

Main Methods:

• Proposed a semantic-enriched augmented graph contrastive learning (SGCD) method.
• Integrated an adaptive denoising technique to handle noisy data.
• Enhanced data through embedding layers and semantic neighborhood mining.

Main Results:

• SGCD effectively mitigates the impact of data sparsity.
• The adaptive denoising method improves model robustness against noise.
• Experiments demonstrated the proposed model's superior performance on real datasets.

Conclusions:

• SGCD offers a robust and effective approach for drug repositioning.
• The method enhances data representation and noise handling capabilities.
• This work advances computational drug discovery by improving existing models.