A powerful, unique tool has been developed to help scientists quickly identify molecules that have the potency to neutralize the SARS-CoV-2 virus, which causes COVID-19, before it actually invades human cells, or totally disabling it before it causes any severe harm.
REDIAL-2020: Online Suite of Computational Models to Screen Molecules
In the paper, "A machine learning platform to estimate anti-SARS-CoV-2 activities," scientists from the University of New Mexico introduced REDIAL-2020, an open-source online suite of computational models designed to assist researchers to swiftly screen small molecules for their COVID-19 fighting capabilities.
According to Dr. Tudor Oprea, chief of the Translational Informatics Division at the UNM School of Medicine, the tool would "replace (laboratory) experiments," adding that it "narrows the field" of what to focus on, thus placing it online for anyone to use.
The UNM team collaborated with scientists from the University of Texas to develop REDIAL 2020 starting last spring after the National Center for Advancing Translational Sciences (NCATS) presented its data from its COVID-19 studies.
Using the NCATS data, the researchers built solid machine learning models, Oprea added.
Screening Novel Compounds for SARS-CoV-2 Fighting Properties
REDIAL-2020, the study further states, is a fast, reliable system to screen novel compounds for their SARS-CoV-2 neutralizing functions, estimating viral activities from its molecular structure. Leveraging NCATS data, researchers developed 11 categorical machine-learning models exposed on the REDIAL-2020 portal to process each submitted molecule for evaluation. After a similarity search, the top ten similar molecules to the query molecule in existing COVID-19 databases and other experimental data are shown. This would allow drug researchers to evaluate the accuracy of the machine learning predictions.
Molecule's Strength to Neutralize Virus Measured
Results from laboratory assays from NCATS measured each molecule's strength to neutralize viral entry, infectiousness, and replication-an ability such as the cytopathic effect or the ability to defend a cell from being waylaid by a virus, a TechXplore article said.
Biomedicine researchers tend to concentrate on positive findings from their work, but in NCATS, scientists also reported molecules with no virus-combatting properties. Such negative data improved the accuracy of machine learning, Oprea added.
The objective was to pinpoint molecules that fit the profile, Oprea said, further saying that his team would want to find molecules that are capable of doing the things we expect them and not those we don't want.
"Multi-Drug Cocktail" to Inhibit SARS-CoV-2
SARS-CoV-2, Oprea noted, was a difficult enemy and stressed no single drug could attack it on its own and said a "multi-drug cocktail" could fight the virus on multiple fronts.
REDIAL-2020 is fueled by machine-learning algorithms that process immense amounts of data while discarding unperceivable hidden patterns, a Technology Networks report said. The UNL team confirmed the machine learning forecasts using the NCATS data to compare with the approved drugs' known effects.
The computational workflow could be developed to assess compounds against other pathogens and further analyze chemicals not yet approved for human consumption.
The research's main intent is to repurpose drugs, Oprea said, but his team is zeroing in on any small molecule. He said the drug needs not be approved, but testing their molecule should yield something significant.
Check out more news and information on the SARS-CoV-2 virus on Science Times.