Molecular-evaluated and explainable drug repurposing for COVID-19 using ensemble knowledge graph embedding.

Area of Science:

• Computational biology
• Drug discovery
• Artificial intelligence in medicine

Background:

• The urgent need for effective COVID-19 treatments persists due to the lack of clinically proven drugs.
• Drug repurposing, identifying new uses for existing drugs, is a popular strategy in pharmaceutical research.

Purpose of the Study:

• To develop a novel drug repurposing approach for COVID-19 utilizing knowledge graph (KG) embeddings.
• To enhance the prediction accuracy of potential COVID-19 therapeutics through advanced computational methods.

Main Methods:

• Constructed a COVID-19-centric knowledge graph and learned "ensemble embeddings" for entities and relations.
• Employed a deep neural network trained on KG embeddings to discover potential COVID-19 drugs.
• Validated predictions using molecular docking and extracted explanatory paths from the KG.

Main Results:

• The proposed method identified more in-trial drugs compared to existing approaches.
• Fosinopril was identified as a potential ligand for the SARS-CoV-2 nsp13 target through molecular docking.
• Explanatory paths from the KG provided reliability and interpretability for the drug repurposing predictions.

Conclusions:

• The KG embedding-based approach offers a reliable and interpretable method for COVID-19 drug repurposing.
• Molecular evaluation and KG-derived explanations enhance the confidence in predicted drug candidates.
• This methodology provides a complementary and reusable framework for assessing KG-based drug discovery.