I met Demis during his previous career in the video games industry. Genuinely lovely guy, insanely gifted and I'm really happy to see him recognised in such a way.
I instantly thought of Futurama's Professor Farnsworth. "This experiment may just win me the nobel prize!" "In what field?" "I don't care. They all pay the same."
The way in which the shape affects the function can be thought of much the same way the shape of a machine affects its function (take for instance a drill vs a sander). Proteins really function a bit like molecular robotics and machinery. They are not static. Very much like an automated production line in a factory, can take a bit of molecular matter, move it around, and then do something with it. Add it to something else, remove part of it, reshape it, bring it into position for the next step, etc. Just so do proteins function. DNA is used as a template for building amino acids and proteins. And the functions of proteins can include anything from structural, through metabolic processes, to DNA manipulation and replication. That is why the shapes are so important.
@@crappymeal What powers which process exactly? Protein folding is largely driven by the attraction/repulsion between different parts of the protein (via coulomb interactions, dispersion forces, additional covalent bonds like disulfide bonds, etc.) and the hydrophobic effect (which itself is mostly driven by the entropy of water molecules).
@@crappymeal The power for the process... Hmm. I'm not so clued up that I can go into it authoritatively at that level, but ATP and KREBs cycle are probably involved somewhere right at the start. Also I recently saw a bit about the way the molecular motor of cilia works, and part of it showed how energy released by the bonding process re-used for the next step. So in the end I think every scrap of energy is taken as far as possible. It's a huge, highly refined system. And it's a super complex topic. Way beyond me for the most part.
To extend an answer to Brady’s question around 4:00, it’s actually quite the opposite-the stuff they are made of is important, but actually the shape of it and how it’s folded is the critical factor, especially if you extend “shape” into other regimes (like the “shape” of charge distribution, for example). In biology this is the principe of “form determines function.” This is well demonstrated with antibodies: when a B cell in your immune system detects a foreign object, it needs to make an antibody where its tips are specially crafted to only grab onto that foreign object in order to “recognize” it in the future. But your body has many B cells in it, and also different bodies have their own B cells-it turns out that two different B cells will grab the same object to identify but make different protein sequences for it, each with the ability to “grab” and detect the foreign object.
This is a really big deal. Protein folding is so complex it would have seemed unthinkable to be able to simulate it only a couple of decades ago. The implications for medicine, pharmaceuticals, manufacturing and so many other fields are limitless. We're a big step closer to being able to design molecules to perform chemical synthesis at will.
Demis Hassabis is also a strong chess player. Someone(I can't recall his name) posted a short video on twitter from a chess + poker night, where Demis was sitting at the poker table with Magnus Carlsen, Hikaru Nakamura and Peter Svidler. Vishy, MVL and Giri was also seen mingling in the room. This event took place just a day or two after the prize was announced.
The structure is important because it *determines* what a protein does and how it behaves chemically. This is obviously true for structural proteins, but also for enzymes and hormones.
I read somewhere that since there is no Nobel prize in biology, there is an unofficial tradition to dedicate the chemistry prize every other year to biology-ish chemistry, and this was such a year. I haven't fact checked this.
I wonder if this will help our understanding of prion disease given that it's so deadly and is caused my proteins within the brain folding up into the incorrect shapes.
There are proteins that are difficult to obtain crystals for, or to obtain high-resolution X-ray diffraction for. AlphaFold makes an interesting starting point for these cases and accelerates the research.
I feel the Pereodic Videos team needs a nobel for popularizing chemistry and making learning chemistry fun and exciting🥳. Thank you so much for the regualr videos.
If machine learning can find a way to fold proteins that enables *me* to have epic hair like Professor Martyn Poliakoff, I'll be as happy as a clam at high tide!
Some additional context on the Professor's description of the winners: John Jumper is a chemist, not just a computer scientist. He got his PhD in chemistry from the University of Chicago in 2017.
It more chemistry than the physics prize was related to physics. To me this, chemistry prize is chemistry while the physics prize is computer science or physiology at the closest to an actual Nobel prize.
Having been part of different "folding at home" groups, but after some of the British groups are no longer functional I haven't done much "folding at home" in a number of years, along with the price of electricity it's not so cheap to leave my PC running all day every day compared to 10 years ago
@@eliasross4576 my core2 quad pulled around 100W while my current ryzen 5 is set about 90W so not much difference between them but price of electricity is about 8x the price of 10+ years ago. And that makes a massive difference
Just like how you could allow seti to run a program on your computer around the 90s/2000s, you can also download a program that runs in the background that computers protein folding. Distributed computing is still a thing you can do now
In 1989 during my biology HS I said that this is the future of protein research and computers will help us model the complex structures of proteins from a chain of aminoacids. Unfortunately at the time it was considered science fiction and impossible
Another 2 nobel prizes for Cambridge university, Dennis Hassabis went to Queens college. But the professor must be happy that another member of King's won a nobel prize even if it was in physics. From another member of King's college. Also did the professor know Geoffrey Hinton while they were at King's, though different years and fields
Ah the Professor is such an innocent. That joke about AI was only one among many but most centred on the fact that the whole large language model thing is basically .. um .. not very real. That is it has no actual intelligence of any kind. (Speaking as an actual Strong AI scientist.) The Chemistry prizes look far more legitimate than the physics ones though - not wondering into that territory.
The topic of teaching is very important to me. I think that at least one year, the Prize should go to some really prolific teacher - one who has taught a lot of chemistry to a very large number of people, and who has made the subject more accessible. Wouldn't you agree?
If it's only 90% accurate, how do we know which are correct and which are wrong? Who tested this? Wouldn't it be more accurate to say it has made a lot of guesses?
@@paulkepshire5056 neural nets don't have hard rules. a domain expert can create filters on the input and output, adjusting the output or sending it back through the neural net, but the neural net itself can not error correct. The closest the combination of neural net, expert filter and user skepticism can get is *plausible*.
I think there's certainly room (and time) for QC to contribute. It's a completely different approach to solving folding problems than ML/AI. I would agree that the panacea of QC didn't really work out (yet) as there were so many more problems with noise and error correction than anticipated.
Chess is just one very basic game, the number of moves is based solely on the number of parts in the board, it is by no means a mark of intelligence, you can get children who beat masters of the game but they have no idea how to make their own lunch
Jinbo at Chicago “invented” the idea that made alphafold successful (applying transformer architecture to multiple sequence alignments; and other learning methods applied to MSAs were demonstrated before alphafold by David’s lab). Google just did it bigger. It is a real bummer that the Nobel committee didn’t recognize that. Gives me doubt on other awards now that I've seen what happens in a field I am familiar with. Alphafold is a great contribution to the field regardless, two people from one paper is over the top though. There was an incredible amount of work that went into protein structure prediction for decades. Disappointing way to recognize this work.
Multiple people get the idea and race to be the first to build a working thing or prove the concept or find the engineering process, that's how it is in science. The details and result matter, and yes unfortunately it goes to those with money to chase results faster. Also, the Nobel is infamous for awarding very few individual people instead of teams. Like the people who found the Higgs Boson but it was only Mr Higgs who got the award, because they couldn't decide on which single person would get it out of the hundreds of people on the team.
You only award Nobel to something that works, not hypotheses. Also using MSA to predict structure is such an old idea (like 40+ years), and everyone in CASP uses it, ideas of adding transformers to it alone is hardly very innovative.
I hope these dimwit geniusi have put together a 4 bazillion dollar endowment to pay for all the medical litigation that's going to be coming down the pike for the next coupla decades, huh?
Would be cool to see one these people show the giants bodies in the us out west in Wyoming there's 3 one probably the one that left the devils tower which is an achilies tendon hexagonal columns banded together each surrounded by a sack of lucien rich protein and u can tell by the wrinklezone at the top of the tower where it stops at a banded section then the aztec serpant gods body on entire east coast
The amount of permutations of how the protein can fold is that high. It is like with lottery numbers: 6 out of 49 doesn't sound much but it is a 9 figure number of probability to actual win the jackpot. A protein is made up of way, way, way more moving parts than 6 out of 49.
@@FutureChaosTV in theory, in reality then it wouldn't exist, it's that simple, you couldn't have a protein made up of every atom in the universe, the statement is still false
@@phonotical Nobody is talking about the number of atoms in the protein. The statement is about the number of possible conformations - the ways that all of the bonds between the atoms can be oriented. A relatively small amount of atoms can lead to a huge number of possible conformations.
I don't think this video will do a service to your institution. If I was to study chemistry in 2024, I'd like professors who know what deep learning is 😂 Don't ignore this stuff.
Well, he's in his 70s now, I doubt he's doing very much teaching anymore. He's old and it's outside of his field so I don't think you can really blame him for that