One of biology's biggest mysteries 'largely solved' by AI

Biskit

Cat whisperer
Supporter
Joined
Nov 23, 2011
Messages
1,321
Location
Sitting in the sun (between the rain storms)
I can't help feeling that the headline is a bit over-egged. I'm not a biologist, but it sounds awfully like the way topics in physics such as quantum mechanics and relativity were introduced when I was a student - "at the end of the nineteenth century physicists thought everything was solved except for a few details..."

Even so, it's a fascinating development - working out how proteins fold using AI.

One of biology's biggest mysteries 'largely solved' by AI
 
Interesting. I think Laurence E. Dahners in Bioterror an Ell Donsaii novel uses this ability as a major plot device. But instead of an A.I. it's a hyper genius human posing as a computer.
 
Much non-scientific journalism either misses the point of the research or or over reports only part of the results.
From what I've read, the new process is about as good as the traditional models, just faster [and cheaper?]. So no real advance in Medicine but another step for Machine Learning.
 
So how long to Jasper Fforde type home sequencing kits and people buying kits and growing a dodo in their garden shed? :D
 
Much non-scientific journalism either misses the point of the research or or over reports only part of the results.
From what I've read, the new process is about as good as the traditional models, just faster [and cheaper?]. So no real advance in Medicine but another step for Machine Learning.
But cheaper and faster does mean it can be a routine tool rather than something which is still "cutting edge" and specialised.
 
Even so, it's a fascinating development - working out how proteins fold using AI.

One of biology's biggest mysteries 'largely solved' by AI

Being a bit pedantic here ;) it doesn't work out how proteins fold (we know on the highest level the protein will 'want' to fold into a configuration that gives the structure the lowest energy possible), it just predicts what shape a protein will take given a specific input.

And to do this it relies completely on a database of hard work by experimentalists over the past years and decades of ~180,000 known examples, see the link in the article.

I am no expert in biology either :), so I have no idea if the known examples are a representive sample of all possible proteins but my understanding is that given that there 20 different ammino acids the number of potential proteins possible from a chain of 100 ammino acids is 20^100. After reading that point it rapidly got very complex! Especially for 10am in the morning. Definitely need another coffee...
 

Similar threads


Back
Top