AI company called DeepMind has essentially cracked the problem of protein folding - AlphaFold2

[HAL9000]

AI company called DeepMind had essentially cracked the problem of protein folding – that is they had managed to successfully predict the 3D structures of complex biochemical molecules by only knowing the 2D sequence of amino acids from which they are made.

Audio 45 seconds to 20 minutes 18 seconds

BBC World Service - Science In Action, Your molecular machinery, now in 3D

DeepMind publishes 350,000 of the structures which make all living things tick.

www.bbc.co.uk

From the program, if scientists understand the shape of a protein it can then help in understanding how the protein functions. It may lead to new drug treatments.

This can also be use for enzymes. One application is plastic. Scientists might be able to find an enzyme that's good at breaking down plastics

Proteins

xkcd.com

Another version of the same story

AI breakthrough could spark medical revolution

A program has been used to predict the structures of nearly every protein in the human body.

www.bbc.co.uk

 

[HAL9000]

If you have a look at explain xkcd

1430: Proteins - explain xkcd

Explain xkcd is a wiki dedicated to explaining the webcomic xkcd. Go figure.

www.explainxkcd.com

it gives you some idea why protein folding is such a hard task

Levinthal's paradox is a thought experiment, also constituting a self-reference in the theory of protein folding. In 1969, Cyrus Levinthal noted that, because of the very large number of degrees of freedom in an unfolded polypeptide chain, the molecule has an astronomical number of possible conformations. For example, a polypeptide of 100 residues will have 99 peptide bonds, and therefore 198 different phi and psi bond angles. If each of these bond angles can be in one of three stable conformations, the protein may misfold into a maximum of 3198 different conformations (including any possible folding redundancy). Therefore, if a protein were to attain its correctly folded configuration by sequentially sampling all the possible conformations, it would require a time longer than the age of the universe to arrive at its correct native conformation. This is true even if conformations are sampled at rapid (nanosecond or picosecond) rates. The "paradox" is that most small proteins fold spontaneously on a millisecond or even microsecond time scale. This paradox is central to computational approaches to protein structure prediction.

 

[Saul Goodman]

Glad I wasted all those years and all that electricity on folding at home, just for some jumped up AI company to steal my thunder.

 

[a_chap]

Scientists might be able to find an enzyme that's good at breaking down plastics

Excellent news! I can see no possible problems with that whatsoever...

 

[SpookyFrank]

This is a big deal for moleular biology if it actually works. When I was a student it was precision gene editing tools (which we now have in crispr) and figuring out protein folding which were the two things that kept coming up in the form 'if we could do this one thing, we could do all these shitloads of other things'.

You could potentially use this to make custom antibodies, or mRNA vaccines to new pathogens, on a very short timescale.

 

[Throbbing Angel]

Heard one of the teams who did this talking on R4 yesterday morning whilst I was on the way to work.

They've given the research/tool/finding/tech away, made it free to use as the Graphene team did. For the common goooooood. Which is great, I think.

 

[HAL9000]

Heard one of the teams who did this talking on R4 yesterday morning whilst I was on the way to work.

They've given the research/tool/finding/tech away, made it free to use as the Graphene team did. For the common goooooood. Which is great, I think.

Google owns DeepMind, they're going to want to make money out of DeepMind. Interesting see what happens next since they've made the code open source.

DeepMind’s revenues, which are derived entirely from applying its technology to commercial Google projects such as making its data centres more energy efficient and improving the voice of its virtual assistant, increased 158 per cent in 2019 to £265.5m in 2019. But turnover was once again dwarfed by pre-tax losses of £460.9m, down 2 per cent on 2018. Administration expenses — which include staff and infrastructure costs — grew 26 per cent to £717m. In addition, DeepMind said that its parent company had “waived the repayment” of loans and interest totalling £1.1bn, which had accumulated over recent years.

Don't know how to link to FT articles, if you want to read it. Search for...

"ft Google’s AI unit DeepMind swallows £1.6bn as losses continue"

 

[HAL9000]

There's another software tool which predicts protein folding, RoseTTAFold.

RoseTTAFold is a “three-track” neural network, meaning it simultaneously considers patterns in protein sequences, how a protein’s amino acids interact with one another, and a protein’s possible three-dimensional structure. In this architecture, one-, two-, and three-dimensional information flows back and forth, allowing the network to collectively reason about the relationship between a protein’s chemical parts and its folded structure.

Video here talks about AlphaFold and RoseTTAFold. If you open the video in youtube, you will see under 'show more', all the information sources they used to create this video.

 

[HAL9000]

Alphafold released the predicted structures for almost every protein known to science – over 200 million structures in total.

AlphaFold has predicted the structures of almost every known protein

Google offshoot DeepMind has released more than 200 million predicted 3D structures, covering nearly the entire protein universe

www.chemistryworld.com

Why protein prediction may not help with drug discovery

Why AlphaFold won’t revolutionise drug discovery

Protein structure prediction is a hard problem, but even harder ones remain

www.chemistryworld.com

Demis Hassabis, founder and CEO of DeepMind, has said he wants a digital simulation of a cell. My guess is this will take some time to generate, but if it can be done, may be that will be another big leap in biology.

 

[HAL9000]