Very cool: researchers use "artificial intelligence" to find a brand-new antibiotic
One of the hot-button concepts among journalists is "artificial intelligence"--AI. Like all new fads, journalists have been writing endless stories about how faaabulous it is, but I've been skeptical. But just read about a great application:
As many of you may know, bacteria are constantly evolving to become more resistant to antibiotics. For many years, researchers stayed ahead of the bacteria, and had at least one new antibiotic that hadn't been used yet--meaning bacteria weren't resistant to it. But today we've lost that margin of safety: some bacteria can't be killed by any drug we have.
So researchers wondered if a clever computer program--AI--could be taught to examine drugs or molecules known to have antibiotic action, identify features that these seemed to have in common, and then scan millions of candidate compounds to find ones that seemed to have these effective characteristics.
The team gave their program descriptions of 6,000 molecules known to have at least some antibiotic effects. The program then compared these, looking for any features these compounds had in common--shapes, charge distribution, acidity or basicity and so on. Once these features were known, the developers had the program scan 1.5 Billion molecules, looking for the properties it had identified as seeming to be effective.
The program found something like ten likely candidates out of 1.5 billion compounds it examined. The one below is called "halicin," and had already been tested several years ago to treat diabetes. It wasn't particularly good at that, and no one tested it for any other use, so it was ignored. But after the computer program flagged it as probably effective as an antibiotic, researchers tested it and it worked.
One interesting thing about this molecule is that it doesn't appear to have features that match any current antibiotic, suggesting it gets its effectiveness by attacking a vulnerability in bacteria that we've never seen. And sure enough, they found it worked by sequestering iron inside bacteria, which wrecks the bacteria's ability to control pH in its cell wall--a mechanism never known before.
The success of this research suggests that the same technique could be expanded to find other drugs effective against other diseases. Very cool.
Anyway, for those curious here's the source.
As many of you may know, bacteria are constantly evolving to become more resistant to antibiotics. For many years, researchers stayed ahead of the bacteria, and had at least one new antibiotic that hadn't been used yet--meaning bacteria weren't resistant to it. But today we've lost that margin of safety: some bacteria can't be killed by any drug we have.
So researchers wondered if a clever computer program--AI--could be taught to examine drugs or molecules known to have antibiotic action, identify features that these seemed to have in common, and then scan millions of candidate compounds to find ones that seemed to have these effective characteristics.
The team gave their program descriptions of 6,000 molecules known to have at least some antibiotic effects. The program then compared these, looking for any features these compounds had in common--shapes, charge distribution, acidity or basicity and so on. Once these features were known, the developers had the program scan 1.5 Billion molecules, looking for the properties it had identified as seeming to be effective.
The program found something like ten likely candidates out of 1.5 billion compounds it examined. The one below is called "halicin," and had already been tested several years ago to treat diabetes. It wasn't particularly good at that, and no one tested it for any other use, so it was ignored. But after the computer program flagged it as probably effective as an antibiotic, researchers tested it and it worked.
One interesting thing about this molecule is that it doesn't appear to have features that match any current antibiotic, suggesting it gets its effectiveness by attacking a vulnerability in bacteria that we've never seen. And sure enough, they found it worked by sequestering iron inside bacteria, which wrecks the bacteria's ability to control pH in its cell wall--a mechanism never known before.
The success of this research suggests that the same technique could be expanded to find other drugs effective against other diseases. Very cool.
Anyway, for those curious here's the source.
posted by SF at
1:49 PM
0 Comments:
Post a Comment
Subscribe to Post Comments [Atom]
<< Home