Protein folding thread

Post Reply
Xyls
Posts: 29
Joined: Sun May 16, 2021 9:20 pm

Protein folding thread

Post by Xyls »

DeepMind says it will release the structure of every protein known to science

https://www.technologyreview.com/2021/0 ... -proteome/

Back in December 2020, DeepMind took the world of biology by surprise when it solved a 50-year grand challenge with AlphaFold, an AI tool that predicts the structure of proteins. Last week the London-based company published full details of that tool and released its source code.

Now the firm has announced that it has used its AI to predict the shapes of nearly every protein in the human body, as well as the shapes of hundreds of thousands of other proteins found in 20 of the most widely studied organisms, including yeast, fruit flies, and mice. The breakthrough could allow biologists from around the world to understand diseases better and develop new drugs.
User avatar
funkervogt
Posts: 159
Joined: Mon May 17, 2021 3:03 pm

Re: Protein folding thread

Post by funkervogt »

People will have a lot of fun for years verifying the accuracy of these.
User avatar
raklian
Posts: 199
Joined: Sun May 16, 2021 4:46 pm
Location: North Carolina

Re: Protein folding thread

Post by raklian »

funkervogt wrote: Fri Jul 23, 2021 12:52 pm People will have a lot of fun for years verifying the accuracy of these.
More likely they will use neural networks or even a later iteration of DeepMind to verify them. :lol:
To know is essentially the same as to not to know. The only thing that occurs is entropy.
User avatar
Yuli Ban
Posts: 1557
Joined: Sun May 16, 2021 4:44 pm

Re: Protein folding thread

Post by Yuli Ban »

AI protein folding breakthrough could spark medical revolution
Artificial intelligence has been used to predict the structures of almost every protein made by the human body.

The development could help supercharge the discovery of new drugs to treat disease, alongside other applications.

Proteins are essential building blocks of living organisms; every cell we have in us is packed with them.

Understanding the shapes of proteins is critical for advancing medicine, but until now, only a fraction of these have been worked out.

Researchers used a program called AlphaFold to predict the structures of 350,000 proteins belonging to humans and other organisms.

The instructions for making human proteins are contained in our genomes - the DNA contained in the nuclei of human cells.

There are around 20,000 of these proteins expressed by the human genome. Collectively, biologists refer to this full complement as the "proteome".

Commenting on the results from AlphaFold, Dr Demis Hassabis, chief executive and co-founder of artificial intelligence company Deep Mind, said: "We believe it's the most complete and accurate picture of the human proteome to date.
And remember my friend, future events such as these will affect you in the future
User avatar
funkervogt
Posts: 159
Joined: Mon May 17, 2021 3:03 pm

Re: Protein folding thread

Post by funkervogt »

Further progress:
I will not attempt to describe the details of this approach - given enough time I think that I probably could, but I'm pretty sure that I don't have enough pork rinds and Mountain Dew on hand to get me through that process, given my current level of knowledge. What I can tell you, though, is that the method uses the primary sequence of the protein, with no reference to existing structures, and it depends strongly on the torsional angles between each residue. Now, that turns into an awful problem very quickly if you just take it as a big long list of individual torsion angles, but this approach uses Frenet-Serret formulas to deal with these things as a curve in torsion space, which appears to be computationally far more tractable. After working through the sequence that way, the problem drops back into Cartesian space for a brush-up with more conventional energy minimization, with what is presumably a solid starting point based on the optimization of the primary backbone geometry.
https://www.science.org/content/blog-po ... prediction
RGN2 outperforms AlphaFold2 and RoseTTAFold (as well as trRosetta) on orphan proteins and is competitive with designed sequences, while achieving up to a 106-fold reduction in compute time. These findings demonstrate the practical and theoretical strengths of protein language models relative to MSAs in structure prediction.
https://www.biorxiv.org/content/10.1101 ... 2.454840v1
Post Reply