Screen 200 million compounds
machine learning discovers hundreds of potential new crown drugs
International war "epidemic" operation
Science and Technology Daily, Beijing, August 13 (Reporter Liu Xia) According to a report on the US Daily Science website on the 12th, US scientists have discovered hundreds of new coronary pneumonia by screening about 200 million chemical substances with a powerful machine learning method. Candidate drugs.
The research leader and professor at the University of California Riverside, Ananda Sankar Ray, explained that this drug discovery platform is a computer algorithm related to artificial intelligence that can predict the activity of drugs through trial and error learning, and its predictive ability It can continue to be improved, "for the systematic discovery of new drugs for the treatment of new coronary pneumonia, such a platform is an important first step."
In the research, team member Joel Kovalevsky used the ligands of 65 human proteins that interact with the new coronavirus protein, and generated machine learning models for each human protein. These models are trained to A new ligand was identified from its 3D structure.
The research team used these machine learning models to screen more than 10 million small molecules from a database containing 200 million chemical substances, and identified compounds that can most effectively target 65 human proteins that interact with the new coronavirus protein. From these compounds, they identified compounds that have been approved by the U.S. Food and Drug Administration (FDA), such as those used in medicines and foods. They also used machine learning models to calculate the toxicity of various compounds, which helped to eliminate potential toxic candidates.
The researchers said that this approach not only allowed them to identify drug candidates with the most significant activity on a single human protein target, but also discovered some chemicals that are expected to inhibit two or more human protein targets.
Lei said: "What excites me most is the compounds that may evaporate, which brings surprises to inhalation therapy."
Researchers believe that traditional methods that rely on cell culture assays are expensive and may take years to test drugs. Compared with them, their machine learning platform has advantages in initially screening a large number of chemicals. Moreover, the platform can not only be used to develop anti-coronary pneumonia drugs, but also accelerate the development of drugs for other diseases.