A current status rumination on the Wiggle power issue.
|Opened on:||Wednesday, March 5, 2014 - 17:51|
|Last modified:||Thursday, March 6, 2014 - 20:36|
After five weeks of random experimentation my conclusion is that I am still at sea, probably because I am missing quite a bit of understanding. This is the way I am seeing things now (feel free to enlighten, chastise, insult me, etc.):
If we idealize in the opening by idealizing or folding on High power, either by hand or by script, we then find that the ideality scores prevents most moves or scripts from gaining any energy ground because the rise in the ideal score, driven by geometry checks( I think), cancels whatever gains we make. This is experienced by folders as “hardening” of the protein. Most importantly then we are in effect no different from Rosetta. So why bother funding the whole foldit experiment at all.
If we start all folding in low power we are essentially at the point the client was prior to NC and old scripts work because the checks are few and the ideality penalties scores are very low therefore allowing energy gains.
So we come to Medium, which increases the geometry checks a little or separates out what was done prior to NC within the old wiggle. It is currently experienced by me and perhaps others as essentially doing nothing that couldn’t be done by the including or re-including it in another power choice, especially Low power. Nevertheless it or some similar compromise seems to be the only possible answer to the problem, or to at least point to it. If you still think there is a benefit from exploring the resource of the human brain as opposed to a Rosetta program, all you can do is improve the ideality slightly. My question then is why not put Medium's function back under Low and just have one wiggle; or perhaps another path would be lowering the ideality penalty score would work better, since both methods don't increase the permutations a folder has to encounter by finding different combinations of wiggle power with CI, etc. And perhaps more importantly, it also might simplify the effort needed to modify our pre-NC scripts to include both ideality penalties and wiggle power variations.