Signaling repurposable drug combinations against COVID-19 by developing the heterogeneous deep herb-graph method

Area of Science:

• Computational biology
• Drug discovery
• Traditional Chinese Medicine

Background:

• Drug repurposing is crucial for identifying new treatments for Coronavirus disease 2019 (COVID-19).
• Existing research often focuses on chemical drugs and single-target approaches, neglecting the complexity of COVID-19.
• There is a need for personalized treatment strategies that consider different stages of SARS-CoV-2 infection.

Purpose of the Study:

• To explore the potential of herbal medicines for COVID-19 treatment through drug repurposing.
• To develop a computational method for identifying specific herb combinations against SARS-CoV-2.
• To address limitations of current drug repurposing strategies by focusing on traditional medicines and multi-target approaches.

Main Methods:

• Constructed heterogeneous graphs of 'Herb-Compound' and 'Compound-Protein' interactions using virtual screening.
• Utilized a metapath-based embedding approach to build a 'Herb-Compound-Protein' heterogeneous network.
• Employed variational graph convolutional networks to generate precision herb combinations for different COVID-19 stages.

Main Results:

• Generated 24 ranking lists of top-10 herbs targeting specific SARS-CoV-2 proteins.
• Identified 20 distinct herb combinations as potential treatments for four COVID-19 infection stages.
• Provided a freely accessible repository for the code and supplementary materials.

Conclusions:

• The study successfully identified potential herbal treatments for COVID-19 using a novel computational framework.
• The proposed method offers a personalized and stage-specific approach to drug repurposing for infectious diseases.
• This research highlights the potential of integrating traditional medicine with advanced computational techniques for future therapeutic development.