Chemistry professor uses artificial intelligence to model potential COVID-19 antiviral inhibitors
By Ben Panko
While physicians and scientists around the world are testing potential treatments for COVID-19, a Carnegie Mellon University chemistry professor is working to model inhibition of viral proteins by molecules with deep neural networks.
"With current widespread shutdowns of businesses and research institutions, experimental drug discovery and development is severely hindered," Assistant Professor Olexandr Isayev said. "Therefore, computations could play a major role in design and discovery of antiviral therapeutics."
Isayev, who joined the faculty of the Mellon College of Science this year, has previously described his research as laying at the "interface of theoretical chemistry, pharmaceutical sciences and computer science." Using neural networks and artificial intelligence, his lab works to design theoretical new molecules for drugs or materials and then analyze them to see if they would be feasible or useful.
When it comes to COVID-19, Isayev is aiming to build up libraries of small molecules that researchers can search through when looking for compounds that could potentially serve as drugs, a process known as virtual screening. "High-quality datasets of molecules for virtual screening will be released immediately for public use," Isayev said.
Earlier this month, Isayev was awarded a grant by the COVID-19 High Performance Computing Consortium, giving his team access to the powerful supercomputing resources available at the Pittsburgh Supercomputing Center that are now being directed at research on the novel coronavirus. Isayev's team is also planning to develop a new drug design platform using artificial intelligence that could greatly optimize and speed up the virtual screening process.
Isayev has also incorporated COVID-19 into his teaching, recently giving his students in the course Special Topics in Computational Chemistry: Machine Learning for Molecular Science a homework assignment that involved using machine learning to predict the potential effectiveness of small molecules that could inhibit the main viral protease. So-called protease inhibitors are a common group of drugs used to treat viruses such as HIV and Hepatitis C, and they are being closely studied for potential applications in the current COVID-19 pandemic.
“The shelter at home order and teaching online by Zoom have severely disrupted the educational process,” Isayev said. “I was thinking of how to comfort students in my class and show them the relevancy of their education and the power of modern computational chemistry methods.”
Therefore, Isayev encouraged his students to apply their models and test the activity of widely discussed drugs against the novel coronavirus. “Surprisingly or not, none of them were predicted as active (at least against this specific target),” he said. “The quest for the cure continues.”