Artificial intelligence has deciphered the structure of virtually every protein known to science, paving the way for the development of new medicines or technologies to tackle global challenges such as famine or pollution.
This is amazing and for me unexpected. I cannot grasp how important this is, not because it is not important, but because it is bigger than my knowledge.
Cheers
Qazulight
Link:
https://www.theguardian.com/technology/2022/jul/28/deepmind-….
Next step is to evaluate that relationship between proteins (or lack thereof) as they affect various human maladies. This, combined with CRISPR could create a quantum jump in medical ability. The other observation is that it relegates doctors to technician status as I’m guessing the human mind is not able to create these solutions without depending on AI that goes beyond their understanding.
Jeff
At the risk of sounding pedantic on a Friday:
Sequencing a protein is fairly straightforward, and has been for years.
The breakthrough here has been in determining the structure i.e. how the chain of amino acids precisely folds itself up into an active protein. The twenty* amino acids have differing physical chemistry, making the final shape of a chain of dozens** far from intuitive or even (until now) approachable.
As a poor analogy: a ball of yarn, made up of short strands of different yarns, all varying in diameter, clinginess, smoothness…and trying to sequence those so a perfectly-shaped keyhole emerges in the final product.
But the remarkably fast leap from “it can’t be done” to “look, we did it…for everything”*** is stunning. Jaw-dropping. Amazing.
As a starter, it will fundamentally rework the basic lab part of pharmaceutical development.
–sutton
*twenty-ish. Trying to keep the pedantry acceptable here
**or hundreds. Even thousands.
***assuming this is actually true, works in real life, etc.
Oh wow, another crux “this changes everything” thread.
AI is just getting going, for better and for worser. Fastening seat belt.
david fb
One part of science has always been guess the answer and then design an experiment to test it. This new technology gives the opportunity to develop better guesses. But in theory. So far unproven.
The alternative is the Edisonian approach. Try everything on the shelf and hope for a winner. Its out of style these days but does result in better patent protection. When a designed agent works you can always argue it was obvious to one skilled in the art.
Once companies routinely screened libraries of materials through various tests to find new leads. One of the best finds is Roundup herbicide.
And it is Deepmind’s grand announcement via multiple outlets that’s being promoted here. Not saying a bit of circumspection might be in order (I know just about enough on the topic to know I might well be wrong)… but it wouldn’t be the first time a company’s press release has made banner headlines that present a misleading picture…
The twenty* amino acids have differing physical chemistry, making the final shape of a chain of dozens** far from intuitive or even (until now) approachable.
…
*twenty-ish. Trying to keep the pedantry acceptable here
**or hundreds. Even thousands.
Mammalian DNA involves a rather smaller number of different proteins, but runs into the millions of them. Per chromosome.
Mammalian DNA involves a rather smaller number of different proteins, but runs into the millions of them. Per chromosome.
Well, no
The following is mostly true. There are (as always in the natural sciences) occasional exceptions
1)“Mammalian DNA” is composed solely of 4 nucleotides: adenine (A), guanine (G), cytosine (C) and thymine (T).
2) a specific triplet of these (e.g. AGT) code for a specific amino acid. As 4x4x4 = 64, there is plenty of room to code for 20 amino acids
3) broadly, each gene codes for a protein.
4) the whole human genome has around 20,000 genes
5) but, there is a whole lot more going on than just DNA → gene → protein. Much of the DNA that doesn’t code for a gene per se has regulatory functions we’re just staring to figure out.
6) then, the transition from (DNA code in the nucleus) to (working protein in the cytoplasm) is the usual set of complexities between theory and practice.
6a) One key component of this is “messenger RNA”, which has recently been in the limelight, as in “mRNA vaccines”
Nonetheless, with respect, “Mammalian DNA involves a rather smaller number of different proteins, but runs into the millions of them. Per chromosome.” just doesn’t make any sense.
–sutton
Small molecule drugs often work by fitting into enzymes formed from those proteins. We know that proteins degrade and stop functioning if the natural folding is disturbed.
Design of better drugs should be possible once you know how they interact with those enzymes and their exact dimensions when folded.