Wiggle Lock on puzzle 625

Case number:845813-993431
Topic:Game: Tools
Opened by:Susume
Status:Open
Type:Bug
Opened on:Friday, August 31, 2012 - 20:34
Last modified:Saturday, February 23, 2013 - 09:17

I have 23 mutable segments on puzzle 625, and the 12th one cannot be moved by bands during wiggle. Even strength 10 bands will not budge it in any direction. This is especially frustrating since there is no move tool on this puzzle. Closing and reopening the client did not help. When I delete residue 91 (the locked one), then residue 90 is locked instead.

Possibly related is the fact that wiggle is extremely slow on this puzzle.

devprev, Win 7

(Fri, 08/31/2012 - 20:34  |  31 comments)


tokens's picture
User offline. Last seen 2 weeks 1 hour ago. Offline
Joined: 11/28/2011

Also wiggle seems to get easily stuck on this puzzle. Even wiggling at CI 0.00 often does nothing. You have to pull the protein to reduce the score about 200 points before wiggle will start working.

Joined: 04/19/2009

I have also seen wiggle "freeze" on this puzzle (Mac OS, devprev)

brow42's picture
User offline. Last seen 2 days 19 hours ago. Offline
Joined: 09/19/2011
Groups: None

I also notice this. My puzzle currently has 3 added residues...83 is wiggle locked. If I delete 83, then 82 becomes wiggle locked. Tested by applying bands to locked segments and noting that wiggle score change is identical to no bands. On previous puzzles, this has made it very difficult to move entire loops. The alternative was pull tool with rubber bands guiding the pull, very coarse. We can use scissors, but there is no move tool.

If I now use the scissors tool on 82, then 84 is wiggle locked. If I then cut 84, then 83 is wiggle locked.

infjamc's picture
User offline. Last seen 2 years 15 weeks ago. Offline
Joined: 02/20/2009
Groups: Contenders

Even bands of strength 10 are not effective for this puzzle... I would be tempted to re-rate this as Priority 1 if it weren't game-breaking in the "program can't run at all" sense.

infjamc's picture
User offline. Last seen 2 years 15 weeks ago. Offline
Joined: 02/20/2009
Groups: Contenders

Oops, there's a typo-- obviously, I meant to say "if it WERE game-breaking in the 'program can't run at all' sense" (it's not).

On a side note, there is a partial workaround: banding to empty space.

Susume's picture
User offline. Last seen 11 hours 52 min ago. Offline
Joined: 10/02/2011

Banding the locked segment to empty space does not move it either. It can be pulled by hand, or rebuilt to a new position. I find the puzzle unplayable, though clearly some people are playing it.

beta_helix's picture
User offline. Last seen 1 day 15 hours ago. Offline
Joined: 05/09/2008
Groups: None

Hopefully we can figure out what is going on before the long Labor Day weekend... but our apologies if we can't!

Joined: 11/10/2007
Groups: Window Group
Status: Open » Closed

I've updated the puzzle, wiggle should be much more responsive now. You'll need to re-login to the game to get the update, and you may get an error the first time you try to start the puzzle, but it should load the next time.

Joined: 05/19/2009
Groups: Contenders
Status: Closed » Open

resolved-> open. I do not experience any improvement despite having restarted a new client and reloading the puzzle (and indeed getting an error the first time reloading). Even with cl - 0.08 wiggle does not move the thing.

Am I missing something ?

Joined: 05/19/2009
Groups: Contenders

What happens to an evo that was saved earlier from the sticky state and is reloaded ?
Will that be sticky still or will that now improve as well ?
Do we need to completely restart the puzzle in a clean track and start all over or can we resume from a known solution with better wiggle ?

Joined: 05/19/2009
Groups: Contenders

I have now closed all clients with 625, reloaded the puzzle and it still will not move unless there is a severe clash.

Joined: 08/29/2010

There is an improvement i saw already, BUT you have to go backwards... if you are stuck then you are also stuck after the update!
Restart the puzzle and it will go easier...

What i saw is:
Improvement of same recipe before the update:
7874 -> 7880
and after the update:
7874 -> 7894 just doing the same things

A little improvement but not really much...

Joined: 05/19/2009
Groups: Contenders

As it is not moving, there is little point in running acripts. I have suspended all clients for this puzzle.

gitwut's picture
User offline. Last seen 11 weeks 5 days ago. Offline
Joined: 05/18/2012
Groups: Contenders

I read the comments for this puzzle and am a bit frustrated. I seriously doubt that much of anything useful will be obtained from the results.

1) Very few players are pursuing the main goal of trying to create bonds to the sugar molecule. They do at first, but find that they get much higher scores by ignoring the molecule altogether.

2) Due to the many bugs seemingly associated with only this puzzle, it is doubtful that any of the results (even if bonds were pursued) will be optimal.

I have a keen interest in wanting this puzzle to be successful. My wife went into septic shock a couple of times over a period of 3 years before finally succumbing to septicemia. I would really like to participate in helping to improve treatments for this condition.

I hope that Foldit will consider doing another sepsis puzzle once the bugs have been worked out. I would suggest too that there be a bonus score added for each bond created to the molecule--enough so that it is more rewarding to do so than not. If that's not feasible, perhaps a requirement for "x" molecule bonds to be present for a score to count.

Joined: 12/06/2008
Groups: Contenders

It's not just you. I hand-worked this puzzle to a respectable score, then ran some mutating scripts on it. The end result was a 7900+ point solution with NO bonds to the sugar molecule.

I wrapped the protein well enough around the sugar that simple manual substitution of sidechains allowed me to make six or seven H-bonds to it. And the score dropped to under 7500. Shaking the sidechains at that point pushed all the bond-makers but one away from the sugar molecule. Wiggling the backbone broke most of the H-bonds, as well.

This makes no sense, unless there are errors in the scoring function. I did the right thing... and got penalized for it. Go figure.

beta_helix's picture
User offline. Last seen 1 day 15 hours ago. Offline
Joined: 05/09/2008
Groups: None

gitwut,

We will indeed do another sepsis puzzle once the bugs have been worked out.
It might take some time, since the issues with this particular puzzle appear to be quite severe, but I promise you that we will not give up on this puzzle.

This is a very difficult problem and we really want all of you to help us (and our collaborators at Harvard) with it, once the puzzle is posted properly.

Thank you all for your patience on this one.

beta_helix's picture
User offline. Last seen 1 day 15 hours ago. Offline
Joined: 05/09/2008
Groups: None

We're bringing the Sepsis puzzle back!
Please help us beta-test it before we post it to main.

http://fold.it/portal/node/992823#comment-21824

bertro's picture
User offline. Last seen 1 day 21 hours ago. Offline
Joined: 05/02/2011
Groups: Beta Folders

Sorry, gitwut, I did NOT mean to -1 your post. I, in fact, agree with you.

Joined: 05/19/2009
Groups: Contenders

And I wanted to correct that -1 and could not keep the damn arrows apart so now it became -2... Grrr..

Joined: 08/29/2010

I also have to say that there is something wrong with the bonding scoring! I arranged some sheets around but they bond rarely just with hard bonds to just 2-3 Points per Bonding! And because of that the position of a sidechain is far more important to the game as the bonding, i think this will also explain the rare bonding to the molecule...

Joined: 11/10/2007
Groups: Window Group

To help determine why the puzzle seems to not be moving much, can someone post:

- name of a shared solution to start from
- the set of steps tried to get the protein to move
- what you would expect to happen, and what did happen

spvincent's picture
User offline. Last seen 14 hours 17 min ago. Offline
Joined: 12/07/2007
Groups: Contenders

In Contenders shared solutions, I've put a couple of solutions.

spv 7828 (start)
spv 7430. After rebuilding 100-104 in 7828. Wiggle All does nothing to improve this latter solution, despite the presence of backbone scores of -76 and such.

HTH

Susume's picture
User offline. Last seen 11 hours 52 min ago. Offline
Joined: 10/02/2011

In Anthropic Dreams shares: solo3_wont_wiggle

This set of steps works as expected:
load share
wiggle at ci 1
protein moves away from sugar as expected (because of the bands to space)
disable bands while wiggling
score climbs over 7000

However, this gets stuck:
load share (or just undo your earlier wiggle)
disable bands
wiggle at ci 1
score gets stuck at 4538

At this point wiggle is stuck - you can enable the bands to space, add or remove bands, move the existing bands, change ci, shake then wiggle, run Blue Fuse, etc. but wiggle will no longer move the protein.
Expected behavior with bands enabled: bands pull protein away from sugar.
Expected with bands disabled: score climbs over 7000
What does happen: wiggle does nothing

You can drag the protein to change its score, and then wiggle will work again.

Joined: 11/10/2007
Groups: Window Group

Thanks for this information. This has helped tracked down what I think is the cause of the problem, so hopefully we'll get a fix out soon.

Joined: 08/24/2011
Groups: Go Science

I agree with gitwut that we really need a score bonus for interactions with the ligand.

Part of the problem here might be that the loop we're working on doesn't actually want to extend in the direction of the ligand. In other words, it's not a bug in the score function, it's a physical reality that extensions of that loop away from the ligand are lower-energy than extensions of the loop that point in the direction we want. That doesn't mean it's impossible to design a loop extension that interacts with the ligand, but it does mean that we're fighting the foldit engine to do so, because all our tools are set to look for the highest score. The way to fix that problem would be a score bonus that counteracts the advantage of loops pointing in the wrong direction.

infjamc's picture
User offline. Last seen 2 years 15 weeks ago. Offline
Joined: 02/20/2009
Groups: Contenders

In that case, let's try something radical:

Score the ligand only and ignore everything else.

Susume's picture
User offline. Last seen 11 hours 52 min ago. Offline
Joined: 10/02/2011

In addition to the difficulty getting the loop to stay close to the ligand, it is difficult (and costly points-wise) to keep polar sidechains next to the ligand so bonds can form. Many times I manually set hydrophilics in the spots adjacent to the sugar, but the mutate function in the scripts I was using kept replacing them with hydrophobes that scored higher because they were nicely hidden by the ligand. If the ligand is a protein, hydrophobes in the interface may make the ligand stick better, but I'm guessing that's not the case with a sugar. When you specifically want hydrogen bonds between the protein and the ligand, foldit's preference for "orange in blue out" makes it hard to keep solutions where those bonds can form. Maybe this could also be addressed with some kind of bonus; I don't know.

Joined: 08/29/2010

maybe an idea for this would be a slider like the clashImportance called bondingImportance where at this puzzle this will should get frozen to the highest (maybe also useful for other puzzles)

Joined: 11/10/2007
Groups: Window Group
Status: Open » Closed

We've found and fixed what is likely the cause of the wiggle lock bug. This will go out in an update shortly.

Joined: 11/10/2007
Groups: Window Group
Status: Closed » Closed

This fix has been released.

marie_s's picture
User offline. Last seen 1 year 20 weeks ago. Offline
Joined: 05/18/2008
Groups: None
Status: Closed » Open

new sepsis testing
So far,
. the ligand's secondary structure is not "M",

. very hard to make bonds with the ligand,

. wiggle can stop even with CI= 0.02

. isosurface crashes the client
UNHANDLED EXCEPTION
1: RaiseException +88 bytes (no line)
2: library_main +28159709 bytes (no line)
3: library_main +28159636 bytes (no line)
4: library_main +5886272 bytes (no line)
5: library_main +5890454 bytes (no line)
6: library_main +3944470 bytes (no line)
7: library_main +3894924 bytes (no line)
8: library_main +2818160 bytes (no line)
9: library_main +4759917 bytes (no line)
10: library_main +3477484 bytes (no line)
11: library_main +5662137 bytes (no line)
12: library_main +3476729 bytes (no line)
13: library_main +5668692 bytes (no line)
14: library_main +5670999 bytes (no line)
15: gapfnScSendMessage +624 bytes (no line)
16: gapfnScSendMessage +2338 bytes (no line)
17: FillRect +272 bytes (no line)
18: CallWindowProcW +27 bytes (no line)
19: wglSwapBuffers +730 bytes (no line)
20: gapfnScSendMessage +624 bytes (no line)
21: gapfnScSendMessage +2338 bytes (no line)
22: LoadStringW +287 bytes (no line)
23: DispatchMessageA +15 bytes (no line)
24: library_main +5676403 bytes (no line)
25: library_main +3469949 bytes (no line)
26: library_main +28757 bytes (no line)
27: library_main +29 bytes (no line)
28: no symbol (no line)
29: no symbol (no line)
30: no symbol (no line)
31: BaseThreadInitThunk +18 bytes (no line)
32: RtlInitializeExceptionChain +99 bytes (no line)
33: RtlInitializeExceptionChain +54 bytes (no line)

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