854 Stinks

Case number:699969-997240
Opened by:Susume
Opened on:Wednesday, March 12, 2014 - 02:59
Last modified:Friday, March 14, 2014 - 22:17

854 has been the most frustrating and pointless puzzle I have ever played. I never got out of the hand work stage, because every single small change I made, to a protein with no bad fragments, resulted in wiggle dropping several hundred points on the fragment filter. No bad fragments, tweak a sheet, still no bad fragments, shake and wiggle, wiggle creates 6 bad fragments. Same with rebuilds. Same with mutates. Over and over and over. The score function completely failed to give any feedback about what changes were making the protein better or worse, because the score changes were always swamped by the fragment pitfalls. The folding landscape that is supposed to be a funnel with local minima turned out to be all holes and no path between them. I knew my fold would not be a prize winner (it was not all helices, for starters), but if the score can't give any feedback at all on better vs. worse, there is no way to make progress.

(Wed, 03/12/2014 - 02:59  |  5 comments)

Joined: 12/06/2008
Groups: Contenders

The slightest changes to a loop also created bad fragments, where none existed, before. Using the remix tool on them was worthless, as no results could be found. I had to manually rebuild until I found something with no penalty, else (most of the time) go back to the structure I had before I made the original change.

wisky's picture
User offline. Last seen 1 year 17 weeks ago. Offline
Joined: 07/13/2011

I found that rebuilding loops with bad fragments into loops that don't have bad fragments almost always slip right back into bad fragments. ******* irritating.

wisky's picture
User offline. Last seen 1 year 17 weeks ago. Offline
Joined: 07/13/2011

Sus, this is exactly what I was getting at with my "core exists" feedback at http://fold.it/portal/node/997222 (sorry for plugging). The filters are broken, and need major work. I just happen (thank god) to not have issues with fragment filter as much as most other players do.

I understand the problem, though. These filters have been broken since they came out, and have only gotten worse since NC. They NEED work. LOTS of them. I think (*shields self from various rotten fruit and vegetables about to be thrown*) filters even have a place in non-mutate/non-design puzzles. However, the way that they are implemented right now is prohibitive, clunky, and really just a pain in the ass to work with.

bkoep's picture
User offline. Last seen 6 hours 18 min ago. Offline
Joined: 11/15/2012
Groups: Foldit Staff

Thanks, Susume—this is good feedback that the Fragment Filter may be prohibitively stringent.

I know the puzzle is closed now, but do you recall whether you were having fragment problems in a particular part of your design? Was it mostly helical or sheet regions that gave you trouble? Were shorter or longer loops easier to work with? Did problematic loops contain a lot of glycine or proline?

Susume's picture
User offline. Last seen 5 hours 52 min ago. Offline
Joined: 10/02/2011

Excellent questions bkoep, and ones that I wish I had asked myself during the puzzle. I played with 854 some more today to try to answer them. I've attached the secondary structure and one of the primary structures that I spent a lot of time being unable to improve. The fragment problems were almost exclusively in the loops, mostly the long loops - had I noticed that early on, I would likely have abandoned the fold (sheet sandwich with a greek key strand order) or made the long loops longer. The sheets were mostly fine and there were no helices. There were no prolines in the loops; the 2-residue loops each had one or two glycines. The long loops did not have any glycines. It's possible wiggle and mutate were gradually expanding the protein from the compact build I started with, forcing those long loops to stretch further than they were able, and that's why they had all good fragments early on but always some bad ones after a few changes. The filters are still annoying and difficult, but the fact that everything I did triggered the fragment filter may have been due to my particular fold.


Developed by: UW Center for Game Science, UW Institute for Protein Design, Northeastern University, Vanderbilt University Meiler Lab, UC Davis
Supported by: DARPA, NSF, NIH, HHMI, Amazon, Microsoft, Adobe, Boehringer Ingelheim, RosettaCommons