Folding Forum

Hi all,

Protein folding is NP-hard, just like the Traveling Salesman Problem, Bin Packing, etc.
Mathematical optimization software (AKA constraint solvers) [1] have been working on ever improving algorithms for such problems for decades.
This software use algorithms such as metaheursitics (tabu search, simulated annealing, late acceptance), linear programming, constraint programming, genetic algorithms, GRASP, branch&bound, ...

The world's top constraint solvers want to help out with improving the algorithms for protein folding for COVID-19:
https://or.stackexchange.com/questions/ ... h-input-da
We want to crowd solve it (think Kaggle).

Can you help us out to get started?

What do we need?

• An input data file, with a data description. Ideally, something that looks like this: https://www.kaggle.com/c/santa-workshop-tour-2019/data but we'll take what's available

• A problem description, including a score function description (= when have we found the ideal protein). Ideally, something that looks like this: https://www.kaggle.com/c/santa-workshop ... evaluation

• An output data file description. Ideally we can send that into folding@home for verification.

[1] https://en.wikipedia.org/wiki/List_of_o ... n_software

If those approaches are applicable at all, you would first have to find a way to map the problem.

F@H works in the time domain, not just finding one configuration but following the "landscape".

The goal isn't really what the final shape or interaction will be. Folding@home builds statistical maps of how likely it is that a protein or molecule will change into a particular shape, then from that shape it tries to find what other shape it can form and how likely are those, and so on. This builds a big map of the entire interactive process. For example, if I'm testing the molecules of a drug to see how they interact with the coronavirus, it's useful to see that 85% of the time the interactions might work along a particular path, but 15% of the time they go in a completely different direction. You need to be able to see that information to get a good understanding of how molecular interactions can happen in the real world. F@h is extremely optimized in building these kinds of maps and statistical pathways.

https://commons.wikimedia.org/wiki/File ... @home.tiff

https://twitter.com/foldingathome/statu ... 3412453378

Interested in helping us develop and enhance the F@h architecture? Come join the development team at our F@h Fireside Chat on Thursday, April 9th, 4-5pm EDT! To receive a Discord invite link, fill out this form: https://tinyurl.com/firesidedev