Hacker News Reader

Comment by Cthulhu_ | Hacker News Reader

What does https://foldingathome.org/ do then? That's been going on for nearly two decades.

Full list of achievements https://foldingathome.org/category/fah-achievements/?lng=en

This is the only real update of the year: https://foldingathome.org/2022/01/03/2021-in-review-and-happ...

SARS-CoV-2 has intricate mechanisms for initiating infection, immune evasion/suppression and replication that depend on the structure and dynamics of its constituent proteins. Many protein structures have been solved, but far less is known about their relevant conformational changes. To address this challenge, over a million citizen scientists banded together through the Folding@home distributed computing project to create the first exascale computer and simulate 0.1 seconds of the viral proteome. Our adaptive sampling simulations predict dramatic opening of the apo spike complex, far beyond that seen experimentally, explaining and predicting the existence of ‘cryptic’ epitopes. Different spike variants modulate the probabilities of open versus closed structures, balancing receptor binding and immune evasion. We also discover dramatic conformational changes across the proteome, which reveal over 50 ‘cryptic’ pockets that expand targeting options for the design of antivirals. All data and models are freely available online, providing a quantitative structural atlas.

Simulating very expensive to compute protein dynamics. These aren't guaranteed solutions, but it's still useful information.

So, even one protein cannot be simulated as in real world?

Even one atom of a heavier element cannot be simulated in the real, depending on what level of detail you want. Multi-atom simulations usually treat them as little non-quantum balls moving around in a force-field that may have been approximated from quantum mechanics.

Not if you're going off of ab initio theory such as Hartee Fock, MP2, CC, etc. We're talking amounts of matrix multiplication that wouldn't be enough to finish calculating this decade, even if you had parallel access to all top 500 supercomputers, you get bigger than a single protein, it's beyond universal time scales with current implementations.

Every time some computer scientist interviews me and shows off their O(n) knowledge (it's always an o(n) solution to a naive o(n**2) problem!) I mention that in the Real World, engineers routinely do O(n**7) calculations (n==number of basis functions) on tiny systems (up to about 50 atoms, maybe 100 now?) and if they'd like to help it would be nice to have better, faster approximations that are n**2 or better. Unfrotunately, the process of going from computer scientist to expert in QM is entirely nontrivial so most of them do ads ML instead

How on Earth? I can't imagine the difference between the computational power of all top 500 supercomputers is THAT many orders of magnitude far off from the computational power of all the folding@home computational power donated by the general public.

supercomputers are specialized products with fast networking to enable real-time updates between nodes. The total node count is limited by the cost of the interconnect to get near-peak performance. You typically run one very large simulation for a long period of simulation time. folding@home doesn't have the luxury of fast networks, jut lots of CPU/GPU. They run many smaller simulations for shorter times, then collect the results and do stats on them.

I looked at the various approaches and sided with folding@home. At one point I had 1 million fast CPU cores running gromacs.

It's not, foldingathome doesn't do those calculations either but uses a simplified model too.

One single iron atom's electrons - 26 of them - contain more degrees of freedom than atoms in the solar system.

A custom supercomputer dedicated to simulating folding proteins (two-state folders with nontrivial secondary and tertiary structure) from unfolded to correctly folded state using only classical force fields probably could work, and DE Shaw has invested a lot of money in that idea: https://en.wikipedia.org/wiki/Anton_(computer)

but, as I pointed out elsewhere, this would not be particularly helpful as it would use an enormous amount of resources to compute something we could probably approximate with a well-trained ML model.

It also wouldn't address questions like biochemistry, enzymatic reactions, and probably wouldn't be able to probe the energetics of interactions accurately enough to do drug discovery.

Nope. Don't dare ask how they treated water/solvation. Lolz.

Now, the question is, does it matter? When do you EVER need to know the exact atomistic, yet alone electronic, trajectory of a single protein starting from a given position within a cell surrounded by waters in a given configuration?

It doesn't really matter. This is the beauty of noise and averages and -- dear to me-- statistical mechanics. At finite temperatures, AKA most everything we experience as living things, quantum details (or precise classical trajectories for that matter) aren't that important for the vast majority of questions we tend to have about a system.

I think it will be extremely useful to be able to simulate cell at real atomic level with all physics involved. That means that you can introduce changes at atomic level in DNA and see its effects. Introduce experimental medications in cell and see its effects. You can actually study cell processes (many of which are not known yet) in simulation. And many, many more. It will be revolution in medicine and in biology in general!

Any atomic level change in DNA is going to have maybe a 1-2 kcal energy difference. To properly calculate that (by running simulations) would require an economically impractical amount of computer time, and doesn't actually really change any of the hard problems in medicine.

Why am I saying this? Because I thought the same as you and it took me 20+ years to realize MD doesn't affect medicine at all.

Reply on news.ycombinator.com