Clashing importance being violated: protein goes through itself

Case number:954892-994476
Topic:Developer Preview
Opened by:Susume
Status:Open
Type:Bug
Opened on:Wednesday, February 6, 2013 - 05:45
Last modified:Saturday, July 20, 2013 - 13:06

I've seen several examples yesterday and today of the protein going through itself at CI values where that should not happen. On 670, while a script was wiggling at .5 ci, the main protein was pulled right through the symmetry ghosts several times (see pic).

On 673, rebuilding at .2 ci, it keeps putting sidechains through the backbone or vice versa (see pic and saved solution "rebuild stuck tryp thru backbone," shared with myself). I expect this at extremely low ci values, but I have been using .2 extensively for many months and have never seen this before.

devprev, Win 7 64

AttachmentSize
foldit_1359960551.png209.73 KB
foldit_1360128939.png399.29 KB
(Wed, 02/06/2013 - 05:45  |  12 comments)


brow42's picture
User offline. Last seen 3 hours 24 min ago. Offline
Joined: 09/19/2011
Groups: None

I have observed this on symmetry puzzles since the first one. Usually it involves banding, if I band the real and mirror parts by hand and wiggle, they may pass through each other (and then explode). I see this at CI 1.0.

bertro's picture
User offline. Last seen 1 week 5 days ago. Offline
Joined: 05/02/2011
Groups: Beta Folders

Exactly to both. I have seen it too.

Joined: 03/18/2012
Groups: Gargleblasters

I have seen this too, but it turned out to be "just" a rather persistent visual glitch. Undoing and redoing, or changing view options, keeps bringing up the crazy solutions, but after saving and reopening the puzzle, the parts were drawn in their reasonable positions.

Susume's picture
User offline. Last seen 2 days 10 hours ago. Offline
Joined: 10/02/2011

On 675 with two bands from protein to shadow, wiggling at ci 1, the backbone got crossed again (see solution "backbone crossed" shared to myself). Earlier with 2 bands to space near the shadow and wiggling at ci .2, it got crossed and wiggle could not get it uncrossed when bands were disabled. We are limited in what we can build if we can't use bands to move the protein close to the shadow.

k2d2 (not verified)
k2d2's picture
Groups: None

You can band or even just nudge and pull the protein (sheets in my particular instance) right through not only the shadow, but through the colored part as well. 675 is also displaying the fairly recently developed behavior of wiggling from +8k to -10k and back in a few seconds... this did not happen in November and December and the first half of January, on any puzzles, now it is frequent.

Even at lowest band setting, trying to bond sheets is not working well. It would be really helpful to be able to band to the catalyst.

After reading the science discussion with Dr. L, I am wondering where the electron pairs come from, especially if using acid for the solvent instead of water.

Susume's picture
User offline. Last seen 2 days 10 hours ago. Offline
Joined: 10/02/2011

In the fixed version of puzzle 675 this problem has disappeared - Thank you!

brow42's picture
User offline. Last seen 3 hours 24 min ago. Offline
Joined: 09/19/2011
Groups: None

In 677, if I do a single weak (0.5) band and pull the monomers towards the origin from a non-interacting distance, wiggle happily mashes them all together with a -9999999 score. It remains there without exploding. Stopping, removing the band, wiggling makes it explode.

If I do the same, but interrupt the wiggle as they come into contact, and periodically start and stop the wiggle, I get a much more sane stopping point with a score around +5k, which then wiggles up to the monomer score when the band is removed.

Susume's picture
User offline. Last seen 2 days 10 hours ago. Offline
Joined: 10/02/2011
Topic: Game: Tools » Developer Preview

This problem has cropped up again in the beta puzzles for 695. See my share to myself in score13 puzzle called "BETA 695 score13 backbone violation 1". It has bands on the main protein only (not the shadows). Wiggle it at full CI and you will see residue 68 go through residue 33 several times(it actually hangs out in contact with it for a while) and residue 59 go through another residue as well. The backbone going through itself should only happen at CI=0. In this case wiggle keeps introducing clashes that take a long time to clear, then exploding, then finally it settles down to a positive score.

tamirh's picture
User offline. Last seen 6 years 10 weeks ago. Offline
Joined: 05/11/2012

As far as I can tell, the other terms in the score function are just out weighing the clashing term. If you are going to be in the scientist chat tomorrow, you can ask in there to get more details about how the scoring function works.

brow42's picture
User offline. Last seen 3 hours 24 min ago. Offline
Joined: 09/19/2011
Groups: None

I have always suspected that on symmetry puzzles, residues move closer together twice as fast as normal, and that causes real and mirror proteins to intersect. As noted in my first comment, symmetry puzzles have always done this. Often I make a second band in the opposite direct to limit motion. I'm so used to it now that I undo and change my banding.

Susume's picture
User offline. Last seen 2 days 10 hours ago. Offline
Joined: 10/02/2011

Interesting to know this has always been the case, and that no one else seems bothered by it. For me it's a serious problem when I plan to have one piece in front of another and it suddenly decides to be behind it (or lodged inside it) instead. I guess I'll have to learn to work around it.

In the current case the real protein was not going through the mirror but going through itself - it was nowhere near the shadows at the time.

[Dang, I hit down arrow when I meant to hit reply - sorry about the down vote - not intentional!]

mimi's picture
User offline. Last seen 28 weeks 6 days ago. Offline
Joined: 11/17/2008
Groups: Contenders

I'm getting this problem today on the 746 ED puzzle using the latest Dev Prev update which I have just loaded. Wiggling loops results in them moving to positions where segments of the protein are intersecting.

Sitemap

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, RosettaCommons