In recent weeks, there have been major advancements in the treatments available for COVID-19 with the UK and Health Canada approvals, and FDA issuance of an Emergency Use Authorization (EUA) of the Pfizer-BioNTech vaccination. Moderna is also nearing approval for their vaccine, and a number of other high-potential therapeutics are making progress in clinical trials.
Outside of the major approvals, there’s also been a global research movement that has taken on the challenge of identifying more effective, long-term treatments for COVID-19. DrugBank continues to contribute by providing researchers with access to a live COVID-19 information dashboard, as well as drug data and clinical trials information to enable advancements in COVID-19 therapeutic research.
Since March, DrugBank has been cited in over 700 publications and research papers exploring potential COVID-19 therapies, including an applied research study led by Mr. Guo-Wei Wei of the Michigan State University’s Department of Mathematics, Department of Biochemistry and Molecular Biology, and Department of Electrical and Computer Engineering. Mr. Wei and his team authored “Repositioning of 8,565 existing drugs for COVID-19,” which was published in June 2020 in The Journal of Physical Chemistry Letters in the American Chemical Society Publications.
Interested to learn more about Mr.Wei’s research, DrugBank reached out for an interview, and summarized his team’s research below.
In early February, Wei’s team of five began research to find a repositioned drug to stop the SARS-CoV-2 virus from replicating and transmitting from person to person. They looked at 26 different proteins in the virus to determine potential targets to inhibit.
Narrowing down the choice of which virus protein to target was decided by three things:
1. The protein should be conservative, meaning it mutates infrequently in comparison to other proteins.
2. The target must be crucial to the virus’ replication. If the inhibitor mutates the virus so it cannot replicate itself, then it cannot transmit from one person to another.
3. The virus protein chosen to inhibit cannot be similar to any human proteins because then the drug will also inhibit the human protein, which can cause genetic diseases.
Their studies showed that the SARS-CoV and SARS-CoV-2 3CL substrate-binding sites are highly conserved and structurally similar, meaning they may bind the same potent protease* inhibitor. Without an effective SARS therapy currently available to start with, Wei’s team referred to reports of X-ray crystal structures and binding affinities of both viruses’ protease inhibitors found in various databases from single-protein experiments.
Wei’s team used artificial intelligence (AI) to help screen candidates from DrugBank—1,553 FDA approved drugs and 7,012 investigational and off-market drugs—to curate the largest available datasets** for SARS-CoV-2 and SARS-CoV main protease inhibitors.
They collected 314 binding affinities for SARS-CoV and SARS-CoV-2 3CL protease inhibitors. From there, they systematically evaluated the binding affinities of all 8,565 drugs found in DrugBank with A machine learning*** model, based on 2D-fingerprinting****, and a 3D pose predictor, called MathPose,29 that the group recently developed themselves to predict 3D binding poses.
“[Without DrugBank], the research would have been miserable,” Wei said. “I would have to go through hundreds, even thousands of papers, and that would take half a year to do.”
Using SARS-CoV-2 3CL protease as the target, Wei’s team predicted binding affinities and provided a top 30 list of potential inhibitors of SARS-CoV-2 from the FDA-approved drugs. Proflavine (DB01123), topped that list for predicted binding affinity, with two hydrogen bonds formed between the drug and the SARS-CoV-2 main protease.
The team also provided a second top 30 list from the 7,012 investigational drugs found on DrugBank, and then narrowed both lists down to a top three of each grouping: FDA-approved and investigational drugs.
The research provided timely guidance for other scientists and narrowed the search for potential antiviral candidates.
“[The paper] caught other peoples’ attention and particularly people who are going to design an experiment and validate to see if this is effective data,” Wei said.
Wei’s team is still working on this research by frequently comparing their new findings to drugs on DrugBank.
Much like the common flu, Wei expects mutations of the virus will require variations of the vaccine. Consequently, he does not feel that one vaccine for COVID-19 will be enough, and plans to continue his research to find long-term solutions.
Wei’s team advocates for the repositioning of existing drugs to quickly find antiviral treatments, but also to provide alternative generics to make more affordable treatments accessible to the larger population.
Are you interested in being featured in DrugBank’s Research Spotlight series? Email us at firstname.lastname@example.org.
*Protease: An enzyme (i.e., protein) that increases the rate of proteolysis, the breakdown of proteins into smaller polypeptides or single amino acids.
**Dataset: A collection of experimental observations of correlated variables, where we would like to use some of them to predict others.
***Machine learning: An application of AI to provide systems the ability to learn from data without explicit programming. The use of algorithms and statistical models to analyze patterns and draw inferences from the data.
****2D Fingerprint: Two-dimensional (2D) chemical fingerprints are widely used in machine learning to predict properties, and as numerical features for the quantification of structural similarity of chemical compounds.