5 ms·

This can't be true, many bioscience people have commented assuring us that because Deep Fold doesn't literally spit out new proteins, its nothing new, and sayin

This can't be true, many bioscience people have commented assuring us that because Deep Fold doesn't literally spit out new proteins, its nothing new, and saying otherwise is misleading

elcomet 4y ago
What do you mean? It certainly can fold new proteins that were not part of its training set, otherwise it is useless. Maybe you mean that it cannot fold proteins in new ways that were never seen in its training set though ? But new proteins might fold in known ways, right?
- bpodgursky 4y ago
  It's sarcasm, every time DeepMind does something revolutionary in protein folding, some nominal biochemist jumps in and says "oh no, because it didn't literally automate 100% of my job today, this is overhyped"
  - TaupeRanger 4y ago
    Well it hasn't improved the human condition one bit yet (except for researchers desperate for paper citations), so whether it lives up to the "hype" remains to be seen.
    - flobosg 4y ago
      DeepMind has a website highlighting some of the applications of AlphaFold (https://unfolded.deepmind.com https://unfolded.deepmind.com), and apparently it has improved at least some lines of research.
      - TaupeRanger 4y ago
        Yes, again very good for the company's PR and padding researcher CVs. But no actual benefit to humankind has happened - again, the hype remains to be justified (or not). Notice the carefully vapid wording: doing [some nebulous thing] to "fight" [some terrible sounding thing], or "this could do [some holy grail amazing thing]".
        flobosg 4y ago
        The same could be said of almost all research, but I see your point. Time will tell, I guess.
        collyw 4y ago
        Agreed, like the James Web telescope, a load of hype about the beautiful images and how much would be discovered as a result of it. Sure we will be able to see further into the universe, but I am still trying to work out what practical application it will have beyond pretty pictures. Anyway, the hype is over, no one in my social media circle who were amazed by it have gone on to post a second picture. Meanwhile a drunk scientist put out a picture of a piece of chorizoand trolled them all. I see it as a merger of "I support the current thing" with "I fucking love science".
        dekhn 4y ago
        actually, tehre are a number of papers published now using alphafold predictions to benefit. No, nobody cured the world of all diseases, but it does represent a sea-change in prediction quality.
    - samus 4y ago
      According to TA, a novel protein nanoparticle is part of some COVID vaccine. It's not far-fetched to assume that such nanoparticles will be designed with machine learning tools in the future.
    - TulliusCicero 4y ago
      And then if that happens, people will say, oh but that innovation was discovered a long time ago, these companies haven't done anything earth-shattering recently.
  - whymauri 4y ago
    I would pay skepticism to entrepreneurial commentaries on AlphaFold (CEO of a company based on competing technology has hot take, who would have guessed?), but I would mind the caveats and areas of improvement pointed out by domain experts i.e. Baker, AlQuraishi, etc. Keeping in mind that the latter usually spend more time praising AlphaFold than not -- and yes, they have more domain knowledge than pretty much anyone on HN: https://moalquraishi.wordpress.com/2021/07/25/the-alphafold2-method-paper-a-fount-of-good-ideas/ https://moalquraishi.wordpress.com/2021/07/25/the-alphafold2... It's almost entirely positive. There was a slice in time where I was effectively paid to find as many gaps in AlphaFold as I could, and, even then, I couldn't help but be impressed.
flobosg 4y ago
Snark aside, this is not AlphaFold per se, but sort of a reverse AlphaFold, the same way protein design can be defined as the inverse protein folding problem.
the__alchemist 4y ago
Tangent: Has AlphaFold and other "ML"/"AI" programs made progress on describing how proteins fold? Or is it solely for final (post-folding) coordinates/dihedral angles? Has there been any progress on understanding folding? Results from folding-at-home? Is it treated as too tough of a problem?
- flobosg 4y ago
  There has been some applications of machine learning in the field of molecular dynamics (and therefore protein folding simulations). Here’s a review from 2020: http://public-files.prbb.org/publicacions/ca66c5f0-0f76-0138-72a9-00155df14f0e.pdf http://public-files.prbb.org/publicacions/ca66c5f0-0f76-0138...
COGlory 4y ago
> This can't be true, many bioscience people have commented assuring us that because Deep Fold doesn't literally spit out new proteins, its nothing new, and saying otherwise is misleading From the article... > But when they instructed microorganisms to make their creations in the labs, none of the 150 designs worked. “They didn’t fold at all: they were just gunk at the bottom of the test tube,” says Baker. For the record, David Baker (and other groups) have been successfully designing new proteins just from their own understanding of protein folding, for years. So "Deep Fold" still has yet to catch up to our own limited human understanding.
- flobosg 4y ago
  Those 150 designs were probably early proofs of concept hallucinated using e.g. RoseTTAFold; Baker has been thinking about protein hallucination for a while already.
- samus 4y ago
  This is for one of the approaches they used. The other approach described in TA had a "decent" (one in ten) success rate.
trash3 4y ago

This can't be true, many bioscience people have commented assuring us that because Deep Fold doesn't literally spit out new proteins, its nothing new, and saying otherwise is misleading

elcomet 4y ago

What do you mean? It certainly can fold new proteins that were not part of its training set, otherwise it is useless. Maybe you mean that it cannot fold proteins in new ways that were never seen in its training set though ? But new proteins might fold in known ways, right?

bpodgursky 4y ago
It's sarcasm, every time DeepMind does something revolutionary in protein folding, some nominal biochemist jumps in and says "oh no, because it didn't literally automate 100% of my job today, this is overhyped"
- TaupeRanger 4y ago
  Well it hasn't improved the human condition one bit yet (except for researchers desperate for paper citations), so whether it lives up to the "hype" remains to be seen.
  - flobosg 4y ago
    DeepMind has a website highlighting some of the applications of AlphaFold (https://unfolded.deepmind.com https://unfolded.deepmind.com), and apparently it has improved at least some lines of research.
    - TaupeRanger 4y ago
      Yes, again very good for the company's PR and padding researcher CVs. But no actual benefit to humankind has happened - again, the hype remains to be justified (or not). Notice the carefully vapid wording: doing [some nebulous thing] to "fight" [some terrible sounding thing], or "this could do [some holy grail amazing thing]".
      - flobosg 4y ago
        The same could be said of almost all research, but I see your point. Time will tell, I guess.
      - collyw 4y ago
        Agreed, like the James Web telescope, a load of hype about the beautiful images and how much would be discovered as a result of it. Sure we will be able to see further into the universe, but I am still trying to work out what practical application it will have beyond pretty pictures. Anyway, the hype is over, no one in my social media circle who were amazed by it have gone on to post a second picture. Meanwhile a drunk scientist put out a picture of a piece of chorizoand trolled them all. I see it as a merger of "I support the current thing" with "I fucking love science".
      - dekhn 4y ago
        actually, tehre are a number of papers published now using alphafold predictions to benefit. No, nobody cured the world of all diseases, but it does represent a sea-change in prediction quality.
  - samus 4y ago
    According to TA, a novel protein nanoparticle is part of some COVID vaccine. It's not far-fetched to assume that such nanoparticles will be designed with machine learning tools in the future.
  - TulliusCicero 4y ago
    And then if that happens, people will say, oh but that innovation was discovered a long time ago, these companies haven't done anything earth-shattering recently.
- whymauri 4y ago
  I would pay skepticism to entrepreneurial commentaries on AlphaFold (CEO of a company based on competing technology has hot take, who would have guessed?), but I would mind the caveats and areas of improvement pointed out by domain experts i.e. Baker, AlQuraishi, etc. Keeping in mind that the latter usually spend more time praising AlphaFold than not -- and yes, they have more domain knowledge than pretty much anyone on HN: https://moalquraishi.wordpress.com/2021/07/25/the-alphafold2-method-paper-a-fount-of-good-ideas/ https://moalquraishi.wordpress.com/2021/07/25/the-alphafold2... It's almost entirely positive. There was a slice in time where I was effectively paid to find as many gaps in AlphaFold as I could, and, even then, I couldn't help but be impressed.

flobosg 4y ago

Snark aside, this is not AlphaFold per se, but sort of a reverse AlphaFold, the same way protein design can be defined as the inverse protein folding problem.

the__alchemist 4y ago

Tangent: Has AlphaFold and other "ML"/"AI" programs made progress on describing how proteins fold? Or is it solely for final (post-folding) coordinates/dihedral angles? Has there been any progress on understanding folding? Results from folding-at-home? Is it treated as too tough of a problem?

flobosg 4y ago
There has been some applications of machine learning in the field of molecular dynamics (and therefore protein folding simulations). Here’s a review from 2020: http://public-files.prbb.org/publicacions/ca66c5f0-0f76-0138-72a9-00155df14f0e.pdf http://public-files.prbb.org/publicacions/ca66c5f0-0f76-0138...

COGlory 4y ago

> This can't be true, many bioscience people have commented assuring us that because Deep Fold doesn't literally spit out new proteins, its nothing new, and saying otherwise is misleading From the article... > But when they instructed microorganisms to make their creations in the labs, none of the 150 designs worked. “They didn’t fold at all: they were just gunk at the bottom of the test tube,” says Baker. For the record, David Baker (and other groups) have been successfully designing new proteins just from their own understanding of protein folding, for years. So "Deep Fold" still has yet to catch up to our own limited human understanding.

flobosg 4y ago
Those 150 designs were probably early proofs of concept hallucinated using e.g. RoseTTAFold; Baker has been thinking about protein hallucination for a while already.
samus 4y ago
This is for one of the approaches they used. The other approach described in TA had a "decent" (one in ten) success rate.

trash3 4y ago