As the author of one such approach, I'm skeptical.
AlphaFold 2 just predicts protein structures. The thing about proteins is that they are often related to each other. If you are trying to predict the structure of a naturally occurring protein, chances are that there are related ones in the dataset of known 3D structures. This makes it much easier for ML. You are (roughly speaking) training on the test set.
However, for drug design, which is what AlphaFold 3 targets, you need to do well on actually novel inputs. It's a completely different use case.
More here: https://olegtrott.substack.com/p/are-alphafolds-new-results-...
That said I'm not sure that's entirely fair, since Alphafold does, as far as I know, work for predicting structures that are far away from structures that have previously been measured.
You're quite wrong about small molecule drug structures. Historically that has been the case but these days many lead structures are made by combinatorial chemistry and are not derived from natural products.
I'm well aware of the impact of natural products and particularly plant secondary metabolites in drug discovery. I'm also aware of combinatorial synthesis occasionally hitting structures that are close to natural products.
But from first principles, why would you need to limit yourself to that subset of molecular space?
Obviously, your structure will need to look vaguely biochemical to be compatible with the bodies chemical environment, but natural products are limited to biochemically feasible syntheses, and are therefore dominated by structures derived from natural amino acids and similar basic biochemical building blocks.
For a concrete example off the top of my head, I'm not aware of any natural diazepines - the structure looks "organic" but biochemistry doesn't often make 7-rings, and those were made long before combinatorial chemistry. Might be wrong on this one, since there's so much out there, but I think it holds.
> very, very, few drugs are "novel" as opposed to being analogues of something naturally in the body
But "analog" means "structural analog" in this context (see https://en.wikipedia.org/wiki/Structural_analog ), which is why people disagreed with you, presumably.
It appears that you were merely saying that ligands must adopt a 3D conformation that's complementary to the receptor. Sure. That's the entire premise of molecular docking software.
But there can be very dissimilar ligands (like morphine and fentanyl) binding the same receptors. A major goal of drug discovery is to find such novel binders, not to regurgitate known ones.
It did very poorly at this last time I checked. Maybe AlphaFold3 is better?
This depends on the application. If you are trying to design new proteins for something, unconstrained by evolution, you may want a method that does well on novel inputs.
> Same with drug design
Not by a long shot. There are maybe on the order of 10,000 known 3D protein-ligand structures. Meanwhile, when doing drug discovery, people scan drug libraries with millions to billions of molecules (using my software, oftentimes). These molecules will be very poorly represented in the training data.
The theoretical chemical space of interest to drug discovery is bigger still, with on the order of 1e60 molecules in it: https://en.wikipedia.org/wiki/Chemical_space
I mean this is a fast award cycle.
[1]: https://www.science.org/doi/10.1126/science.abj8754
[2]: https://cen.acs.org/analytical-chemistry/structural-biology/...
I remember when computer aided drug design first came out (and several “quantum jumps” along the way). While useful they failed often at the most important cases.
New drugs tend to be developed in spaces we know very little about. Thus there is nothing useful for AI to be trained on.
Nothing quite like hearing from the computational scientist “if you make this one change it will improve binding by 1000x”. Then spending 3 weeks making it to find out it actually binds worse.
It needed Oriol as well doing IC work
Also I really hope the Nobel Prize of Economics goes to Bill Gates! He facilitated sooo much advances by releasing Excel that this must be recognized!
And based on this year's announcements so far I am not sure that my sarcastic comments should be taken as a joke!
"These authors contributed equally: John Jumper, Richard Evans, Alexander Pritzel, Tim Green, Michael Figurnov, Olaf Ronneberger, Kathryn Tunyasuvunakool, Russ Bates, Augustin Žídek, Anna Potapenko, Alex Bridgland, Clemens Meyer, Simon A. A. Kohl, Andrew J. Ballard, Andrew Cowie, Bernardino Romera-Paredes, Stanislav Nikolov, Rishub Jain, Demis Hassabis"