Free AI protein software packages nearly predicted structure of the Omicron coronavirus variant correctly
- Reference: 1641904387
- News link: https://www.theregister.co.uk/2022/01/11/ai_in_brief/
- Source link:
Colby Ford, a researcher at the University of North Carolina, ran simulations using DeepMind's AlphaFold and the University of Washington's RoseTTAFold to predict the protein structure of the latest strain dominating current COVID-19 cases.
"One of Ford's two predicted structures proved to be pretty much right: He calculated that the positions of its central atoms differ by around half an angstrom, about the radius of a hydrogen atom," Wired [1]first reported .
[2]
Ford arrived at his conclusions before scientists were able to fully study a real sample of Omicron under an electron microscope and map its structure properly. The variant contains more than 30 changes to the spike protein — the SARS-CoV-2 protein that recognizes host cells and is the main target of the body's immune responses.
[3]
[4]
Although protein prediction models can help speed up Omicron research, having access to a real sample still beats computer models. Scientists were able to figure out that the new variant binds to its host's cells more strongly than previous strains, something AI models couldn't predict.
"The gold standard will always be direct measurement," said Sriram Subramaniam, a professor at the University of British Columbia who studied Omicron samples. "If you're building a billion-dollar drug program, people want to know what's the real thing."
IBM Watson Health is up for sale
IBM wants to sell Watson Health for more than $1bn as it tries to offload its reportedly unprofitable AI medical data and analytics unit.
Industry talk that Big Blue wants to cut its IBM Watson services down are circling again. Last year the idea was floated in the Wall Street Journal ; this time it has popped up in [5]Axios .
[6]
Investment bank BofA Securities is trying to help IBM find a buyer, and bids were due this week. Potential interested businesses supposedly include health companies and private equity firms.
[7]Prisons transcribe private phone calls with inmates using speech-to-text AI
[8]AI surveillance software increasingly used to make sure contract lawyers are doing their jobs at home
[9]Alphabet launches new AI drug discovery startup, Isomorphic Labs, led by DeepMind CEO
[10]Twitter's algos favour tweets from conservatives over liberals because they generate more outrage online – study
IBM has had difficulty trying to make money from Watson Health. It's hard to scale up medical solutions to treat diseases in real-life clinical settings across different patient populations and hospitals.
How AI could stop the production of psychoactive drugs
Scientists have developed an AI model dubbed DarkNPS that's capable of predicting the chemical structure of psychoactive drugs. They believe it could help officials crack down on upcoming narcotics before they're even developed or sold.
"Governments could go through the cache of hypothetical drugs DarkNPS developed and ban them, even before anyone actually produces or distributes them," David Wishart, a computer scientist and biologist at the University of Alberta, [11]told Scientific American in an article published last week.
The algorithm generated 8.9 million potential drug structures from the psychoactive substances it was trained on; nearly 200 designer drugs that have lately emerged on illicit markets were present in this generated set, according to a [12]press release . Given just the mass of a molecule and mass spectrometry data, it was able to more or less predict a drug's chemical formula and structure. ®
Get our [13]Tech Resources
[1] https://www.wired.com/story/ai-software-nearly-predicted-omicrons-tricky-structure/
[2] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=2&c=2Yd23uF@-3vanNu6P6-iH4wAAAEk&t=ct%3Dns%26unitnum%3D2%26raptor%3Dcondor%26pos%3Dtop%26test%3D0
[3] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44Yd23uF@-3vanNu6P6-iH4wAAAEk&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[4] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33Yd23uF@-3vanNu6P6-iH4wAAAEk&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[5] https://www.axios.com/ibm-tries-to-sell-watson-health-again-82f691a4-ab81-4b2b-a5bb-13a7556c8ef1.html
[6] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44Yd23uF@-3vanNu6P6-iH4wAAAEk&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[7] https://www.theregister.com/2021/12/04/in_brief_ai/
[8] https://www.theregister.com/2021/11/20/in_brief_ai/
[9] https://www.theregister.com/2021/11/08/in_brief_ai/
[10] https://www.theregister.com/2021/10/30/in_brief_ai/
[11] https://www.scientificamerican.com/article/how-ai-could-prevent-the-development-of-new-illicit-drugs/
[12] https://news.ubc.ca/2021/11/15/ubc-researchers-train-computers-to-predict-the-next-designer-drugs/
[13] https://whitepapers.theregister.com/
Re: Nice to see AI working
As structure predictions go this isn't hugely hard, it's an already well-characterised protein with just a few residues differing. It's good to be able to do it though.
Re: Nice to see AI working
Quite so, but AI has had some notorious slip ups where small changes in the input data results in large changes in the predictions. It would have been a huge surprise if the omicron spike protein structure had been significantly different from the previous variants, and a bad AI prediction would have been a definite egg-on-face moment for the technique.
AI has great potential for use as a pre-screen if its predictions correlate reliably with experimental data, so that experimental resources, always too scarce, can be concentrated on good candidates but the hard work is ahead on establishing quality and reliability measures for the predictions.
Re: Nice to see AI working
I think the real advantage is that it's another part of the puzzle. The money that has been thrown at SARS-COV2 itself is ludicrous but the tools that have been developed including the sequencing, folding and the analysis of the method of attack and distribution, mean that we're now getting closer to reasonable simulations. This could come in very handy given the size of the list of zoonotic candidates.
Re: Nice to see AI working
But had this research been done in China then lots of idiots would be running around screaming.
They do that all day regardless. Might have briefly changed what they were screaming about, though.
"Governments could go through the cache of hypothetical drugs DarkNPS developed and ban them, even before anyone actually produces or distributes them,"
How does a ban before they are even made stop anyone anymore than the bans on existing illegal drugs?
If anything, this program will point out to suppliers what to aim for in their next 'designer high'.
AFAIK, banning synthetics has been a game of whack-a-mole. Some synthetic gets banned and a shady supplier creates a slightly different drug that's technically legal to sell. No effect on the guy selling on a street corner. The bans have an affect on the type of stores that sell bongs "for tobacco only".
Nice to see AI working
I think that life is improving, these days Omicron is busy vaccinating all the un-vaccinated people. But had this research been done in China then lots of idiots would be running around screaming.
The basis for this AI work has been around for a long time and it's nice to see it working, it has the potential to help a lot of things, my initial thought is what this might do to help human health by documented the good food that we don't like, and bad food that we eat a lot.