AI nearly predicted structure of Omicron coronavirus variant • The Register


In short Using two different free protein prediction AI algorithms, computer scientists were almost able to model Omicron before the coronavirus variant was physically mapped.

Colby Ford, a researcher at the University of North Carolina, performed simulations using AlphaFold from DeepMind and RoseTTAFold from the University of Washington to predict the protein structure of the latest strain dominating current cases of COVID-19.

“One of the two structures Ford predicted turned out to be about right: he calculated that the positions of its central atoms differ by about half an angstrom, about the radius of a hydrogen atom.” , Wired first reported.

Ford came to his conclusions before scientists were able to fully study an actual sample of Omicron under an electron microscope and correctly map its structure. The variant contains over 30 modifications of the spike protein – the SARS-CoV-2 protein which recognizes host cells and is the primary target for the body’s immune responses.

While protein prediction models can help speed up Omicron research, having access to an actual sample still beats computer models. Scientists were able to find that the new variant binds more strongly to its host’s cells than previous strains, which AI models couldn’t predict.

“The gold standard will always be the direct measurement,” said Sriram Subramaniam, a professor at the University of British Columbia who has studied the Omicron samples. “If you are building a billion dollar drug program, people want to know what the real thing is.”

IBM Watson Health is for sale

IBM wants to sell Watson Health for more than $ 1 billion as it tries to offload its supposedly unprofitable AI medical data and analysis unit.

Industry rumors that Big Blue wants to scale back IBM Watson services are circulating once again. Last year the idea was launched in the the Wall Street newspaper; this time he appeared in Axios.

Investment bank BofA Securities is trying to help IBM find a buyer, and the offers were expected this week. Potentially interested companies are expected to include healthcare companies and private equity firms.

IBM struggled to make money with Watson Health. It is difficult to extend medical solutions to treat disease in real clinical settings across different patient populations and hospitals.

How AI could stop the production of psychoactive drugs

Scientists have developed an AI model called DarkNPS, capable of predicting the chemical structure of psychoactive drugs. They believe it could help authorities crack down on upcoming narcotics even before they are developed or sold.

“Governments could dig into the cache of hypothetical drugs developed by DarkNPS and ban them, even before anyone produces or distributes them,” said David Wishart, computer scientist and biologist at the University of Alberta. American scientist in an article published last week.

The algorithm generated 8.9 million potential drug structures from the psychoactive substances on which it was trained; Nearly 200 designer drugs recently appeared on the illicit markets were present in this generated set, according to a press release. Considering the mass of a molecule and mass spectrometry data, it was more or less able to predict the chemical formula and structure of a drug. ®


About Author

Comments are closed.