Why proteins fold and how GPUs help us fold
aval.bearblog.dev
Why proteins fold and how GPUs help us fold
1–10 of 36 posts
Re: Why proteins fold and how GPUs help us fold
#2Was this text AI-generated?
Re: Why proteins fold and how GPUs help us fold
#3Did AlphaFold not use TPUs?
Re: Why proteins fold and how GPUs help us fold
#4I got about a page in before finding out this is drivel. The final straw was "AI companies showed up and solved it in an afternoon". No faster way to show you don't know what you're talking about.
Re: Why proteins fold and how GPUs help us fold
#5I really appreciated the explanation of what proteins are, in simple terms. I assume (?) it's accurate enough for a layperson.
And I do love the optimism.
But then you must admit this reads like a B-movie intro:
Then AI companies showed up in 2020 and said "we got this" and
solved it in an afternoon. ... We're playing God with molecules
and it's working.Re: Why proteins fold and how GPUs help us fold
#6The secondary structure graphic is entirely wrong. It's full of bad chemical formulas, and I would assume is AI-generated.
I'm quite impressed by the amino acid overview graphic. I'm sure all images are AI-generated, and this one is something I didn't expect AI to be able to do yet. There are mistakes in there (e.g. Threenine instead of Threonine, charged amino groups for some amino acids), but it doesn't look immediately wrong. Though I haven't needed to know the chemical formular for all the amino acids in a long time, so there are probably more errors in there I didn't immediately notice. The angles and lengths of the bonds are not entirely consistent, but that also happens without AI sometimes if someone doesn't know the drawing tools well. The labels are probably the clearest indicator, because they are partly wrong and they are not consistent as they also include the non-side-chain parts sometimes, which doesn't make sense.
The biology part of the text looks somehwat reasonable overall, I didn't notice any completely outrageous statements at a quick glance. Though I don't like the "folding is reproducible" statement as that is a huge oversimplification. Proteins do misfold, and there is an entire apparatus in the cells to handle those cases and clean them up.
Re: Why proteins fold and how GPUs help us fold
#7One protein fold is cute.
How many H100s do you need to simulate one human cell? Probably more than the universe can power.
Re: Why proteins fold and how GPUs help us fold
#8Start reading from the 3/4 mark, that’s the ‘how’ part
Re: Why proteins fold and how GPUs help us fold
#9I got about a page in before finding out this is drivel. The final straw was "AI companies showed up and solved it in an afternoon". No faster way to show you don't know what you're talking about.
I don't know the space, so I found the article interesting. Please explain, what's wrong with it?
Re: Why proteins fold and how GPUs help us fold
#10Great article!
On a sidenote, what is this new style of writing using small sentences where each sentence is supposed to be a punchline?
"And most of those sequences? They don't fold into anything useful. They're junk. They aggregate into clumps. They get degraded by cellular quality control. Only a TINY fraction of possible sequences fold into stable, functional proteins."