The Nobel committee usually prefers to wait and evaluate longer-term impact, so I'd be quite surprised. CRISPR was obviously revolutionary in 2013 (imo, more than alphafold), and won the Nobel in 2020.
> more than half a million researchers have used the machine-learning system, generating thousands of papers
Funny things is how the general scientific community (including nature) defines 'impact'. I somehow still strangley trust the Nobel committee to take a different approach here. Was curious and found this interesting collection of references: https://www.researchgate.net/project/Enacting-Excellency-Awa... .
AlphaFold doesn't solve folding. It makes metaheuristic guesses without writing a bunch of quantum chemistry, statistical physics, thermodyanamics, and topology maths / algorithms.
I don't mean to downplay AlphaFold, but we haven't solved protein folding yet. This press is really getting ahead of itself.
Are you a trained scientist, especially in the biochem/medsc field? What makes you more certified to say this isn't a revolutionary tool?
I personally know several people that do research that have unlocked new possibilities through this tool. My wife is a neuroscientist and she's used this tool a few times for reasons that are above my head (even with a Msc in Microbiology). This type of work used to take a PhD student 4 years or more to do a single relatively simple protein. Getting answers within a few seconds is revolutionary.
This reminds me of the criticisms of Copilot (GPT-3 application to generate computer code).
Many engineers will say it doesn't code. It just regurgitates and remixes the data it was trained on. It just makes "meta-heuristic guesses."
But anyone taking an honest and objective view of it can see that Copilot does add value. It's no substitute for a real, human, engineer, but it clearly adds value.
I don't think AlphaFold would get to this level of funding, resource commitment, etc. if it was adding 0 value.
Being a domain expert, I'm curious what value, if any, you think a large transformer model could add to the domain of protein folding. Is it really zero value, in your view?
The computer doesn't solve problems on its own, but it is objectively a breakthrough technology. It's pretty clear that AlphaFold is a phenomenal innovation.
GitHub ripping off everyone’s code to build copilot has caused a small (not as big as it should have been imho) exodus of open source projects to Gitlab and others.
I still don’t see how it’s a good idea to import unknown code with unknown licenses into your project, but apparently if copilot does it it’s okay.
> The AI effect occurs when onlookers discount the behavior of an artificial intelligence program by arguing that it is not real intelligence.
> Author Pamela McCorduck writes: "It's part of the history of the field of artificial intelligence that every time somebody figured out how to make a computer do something—play good checkers, solve simple but relatively informal problems—there was a chorus of critics to say, 'that's not thinking'." Researcher Rodney Brooks complains: "Every time we figure out a piece of it, it stops being magical; we say, 'Oh, that's just a computation.'"
Solving physics isn't a soft thing like making pretty art.
I'm firmly in the "AI/ML will eat the world" camp, but the praises being foisted upon AlphaFold are borderline damaging to the real field and its practice.
You can't throw AlphaFold at pharmaceutical problems and call it a day. This press feels like a "mission accomplished" victory lap when it's very clear we're only just getting started.
I have a Msc in microbio, not biochem, but my understanding is that proteins don't have a constant shape. They vibrate, interact with other molecules, etc.
You won't have a perfect answer unless you want to predict its shape in a vacuum, which wouldn't be very useful either way. Having it "close enough" is already extremely useful. There are definitely edge cases where it gets it wrong, but there are always edge cases in ML. More data = better results with the same architecture.
Tons of things that won Nobel prizes weren't 100% accurate, it's not a prize for solving science, rather a prize for advancing science.
My wife is a neurobiologist and the impact of this advance is groundbreaking for her work.
The Norwegian Nobel Committee which selects the recipients of the peace prize doesn't have anything to do with the physics and chemistry prizes which are awarded by the Royal Swedish Academy of Sciences.
Thanks for supplying the extra detail as a counterpoint to my flippancy. I was aware that the Nobel Prize for Economics was "not a real Nobel Prize" but didn't know the Peace Prize was also quite separate from the science based awards.
It does make sense though, seeing as the scientific awards are generally awarded based on actual breakthroughs, whereas the political ones are, let's say, fuzzier.
> didn't know the Peace Prize was also quite separate from the science based awards.
It's a bit more complicated than that. The committees for Physics and Chemistry (and Economics) are colocated at the Royal Swedish Academy of Sciences. Medicine is elsewhere, as is Literature. Physics, Chemistry, Medicine, and Literature work together for final approval. Peace is completely on its own.
From my chat with friends who work in the area of drug design, AlphaFold is accurate for overall structure, but is not that accurate for predicting structure around interaction locations.
Isn't that quite a big claim? My question, Is that work with Alphafold that significant that it warrants the eyes of Nobel committee? Genuinely curious.
In my opinion, absolutely. The "protein folding problem" has been widely regarded as one of the biggest challenges in molecular biology for over half a century, and Alphafold has effectively solved it. I would put this up there with Sanger winning the prize for discovering how to sequence DNA and Kary Mullis for inventing PCR... this will have widespread implications for allowing us to understand, and even design proteins.
But they didn't solve the protein folding problem. They solved a simpler problem, protein structure prediction.
What is important about their discovery is that we now know for certain that a judicious combination of expensive-to-obtain structure information, and easy-to-obtain protein sequence relationships can be used to build a generalized protein structure predictor (it can predict structures with no prior example of a fold, although there are limits)... and you don't have solve the general folding problem to do it. You do not need to know the path, to get to the destination!
Many of us in the field expected this to be true but there wasn't any really good example to point to that was widely accepted by the community. And in the ~year or so since this was demonstrated, the community has already found a wide range of uses for this that have validated the structure predictions and demonstrated their utility- using open codes and models.
The Nobel Prize does not only award scientists for enabling high impact discoveries, but occasionally to people who make a major discovery that has no immediate impact. There is literature dribbling out from folks using AlphaFold models, but that's not what they would be awarding here. This was a long-standing problem that was convincingly solved.
Not Nature but Molecular Systems Biology [2], apologies, in the context of reverse docking for antibiotics discovery:
"We confirm extensive promiscuity, but find that the average area under the receiver operating characteristic curve (auROC) is 0.48, indicating weak model performance."
Derek Lowe had a post about this earlier this month [1] (which includes important qualifications I failed to omit).
[1] "Benchmarking AlphaFold-enabled molecular docking predictions for antibiotic discovery", Molecular Systems Biology 2022 18:e11081 https://doi.org/10.15252/msb.202211081
This is already known. Structural models are usually not well suited for protein-ligand docking, where proper pocket modelling and accurate side chain positioning are key for the determination of true positive hits.
Sorry, you don't know what you're talking about. Nobody is truly claiming these models can be used directly for drug discovery (well, some people claim that; but they're wrong). We already knew that, though- the same problem exists with high quality crystal structures.
What would be more interesting is if we did a whole bunch of crystal or cryo-EM structures far from what we've previously determined and demonstrate whether alphafold could make out-of-class predictions for them.