I was wondering if anyone has ever tried applying PVS/MTD(f)/Alpha-beta type searcher to the whole 'protein folding' problem? Really it all seems so similar to me. Going depth first, string some amino acids together using a 'legal move' generator based on physics, evaluate the resulting position based on finely tuned evaluation parameters, and then let the 'move ordering'/pruning/extensions tricks begin.
Might vastly improve the current accuracy over the Monte Carlo stuff folding@home appears to do. Spend some of that CPU horsepower tuning the parameters instead. Perhaps that is what Evo@home is doing...I don't know, but I am currently inquiring.
Chess vs. Protein Folding
Moderators: hgm, Rebel, chrisw
-
FlavusSnow
- Posts: 89
- Joined: Thu Apr 01, 2010 5:28 am
- Location: Omaha, NE
-
hgm
- Posts: 27808
- Joined: Fri Mar 10, 2006 10:06 am
- Location: Amsterdam
- Full name: H G Muller
Re: Chess vs. Protein Folding
I don't see the similarity. Chess is a zero-sum game where two independent players strive for exactly opposite aims. Hence you have alpha-beta pruning.
Protein folding is a one-sided optimization problem. No such thing as alpha-beta pruning there. It is in fact a lot like playing golf, if it were not for the flag. The hole is small, and the greens are vast, and the rough even larger. Making the hole deeper doesn't help a thing in getting the ball in it, What would you prune?
Protein folding is a one-sided optimization problem. No such thing as alpha-beta pruning there. It is in fact a lot like playing golf, if it were not for the flag. The hole is small, and the greens are vast, and the rough even larger. Making the hole deeper doesn't help a thing in getting the ball in it, What would you prune?