A network representation approach for COVID-19 drug recommendation

Area of Science:

• Computational biology
• Drug discovery
• Virology

Background:

• The COVID-19 pandemic has caused millions of deaths globally, with no approved antiviral treatments.
• Drug repositioning is a promising strategy for rapid therapeutic development against emerging infectious diseases like COVID-19.
• Existing drug repositioning efforts for COVID-19 are hampered by limited data and the complexity of the virus.

Purpose of the Study:

• To propose a novel computational framework (COVDR) for effective drug repositioning targeting COVID-19.
• To overcome data sparsity challenges in identifying potential COVID-19 therapeutics.
• To identify and validate novel drug candidates against SARS-CoV-2.

Main Methods:

• Development of the COVDR framework utilizing local graph aggregating representation in a heterogeneous network.
• Aggregation of multi-hop neighbors within the heterogeneous graph to maximize the recall of potential COVID-19 drugs.
• In silico validation using molecular docking simulations to assess drug efficacy.

Main Results:

• The COVDR framework significantly outperforms existing baseline methods in COVID-19 drug repositioning.
• Experimental results demonstrate the framework's ability to address data sparsity effectively.
• Molecular docking simulations confirmed the therapeutic potential of three identified drug candidates against COVID-19.

Conclusions:

• The COVDR framework offers a robust and efficient approach for drug repositioning in the context of COVID-19.
• The identified drug candidates warrant further investigation for clinical development.
• This study highlights the potential of graph-based machine learning in accelerating drug discovery for viral diseases.