Network machine learning maps phytochemically rich "Hyperfoods" to fight COVID-19

Area of Science:

• Computational Biology
• Bioinformatics
• Network Medicine

Background:

• The SARS-CoV-2 virus, responsible for COVID-19, interacts with host genes and proteins.
• Identifying novel therapeutic agents, including those from natural sources, is crucial for combating the pandemic.

Purpose of the Study:

• To develop and validate a network machine learning approach for identifying potential anti-COVID-19 molecules from food sources.
• To predict the efficacy of these molecules by targeting the SARS-CoV-2-host interactome.

Main Methods:

• Utilized a supercomputing platform (DreamLab App) to harness distributed smartphone computational power.
• Calibrated machine learning models using a dataset of 5658 drugs targeting COVID-19 interactomics, achieving 80-85% accuracy.
• Screened a database of 7694 bioactive food molecules against the calibrated models.

Main Results:

• Successfully predicted drug repurposing candidates, including statins and metformin.
• Identified 52 bioactive food molecules (flavonoids, terpenoids, etc.) predicted to target the SARS-CoV-2-host interactome.
• Constructed a "food map" illustrating the theoretical anti-COVID-19 potential of food ingredients.

Conclusions:

• The in silico food map provides a valuable resource for precision nutrition interventions against COVID-19.
• This network machine learning method demonstrates potential for discovering natural compounds with antiviral properties.
• Further clinical studies are warranted to validate the predicted anti-COVID-19 effects of identified food molecules.