721 wiggle sticks too

Case number:845813-995243
Topic:Game: Tools
Opened by:jamiexq
Status:Closed
Type:Bug
Opened on:Wednesday, May 29, 2013 - 16:47
Last modified:Wednesday, June 12, 2013 - 11:05

721 wiggle all is sticking - you can wiggle at CI ZERO and the unfrozen part does not move - it should crumple

Using devprev Win7

(Wed, 05/29/2013 - 16:47  |  16 comments)


beta_helix's picture
User offline. Last seen 5 days 18 hours ago. Offline
Joined: 05/09/2008
Groups: None
Status: Open » Open

Can you share your solution with us where this happens?

When I reset the puzzle (using Windows) and wiggle all with CI=0 the helix crumbles onto itself:

jamiexq's picture
User offline. Last seen 16 hours 8 min ago. Offline
Joined: 07/31/2011

I uploaded for scientists score 8831 - I deleted a seg at residue 45 I think

jamiexq's picture
User offline. Last seen 16 hours 8 min ago. Offline
Joined: 07/31/2011

After you mutate at all, wiggle at zero does not move it
I do have "Enable backbone constraints" checked as well

beta_helix's picture
User offline. Last seen 5 days 18 hours ago. Offline
Joined: 05/09/2008
Groups: None

Thanks jamiexq.
Have you noticed this problem on any non-design puzzles?

I know 720 is an issue, what about the Flu Design Puzzle 718?

Have you ever run into this with Prediction Puzzle 719?

jamiexq's picture
User offline. Last seen 16 hours 8 min ago. Offline
Joined: 07/31/2011

Hi Beta, No this is not at all an issue on the flu puzzle 718 - that puzzle is fine
Nor is it a problem on 719 or any but these last 2 that I have seen

I am assuming it has to do with the filters

Joined: 09/21/2011
Groups: Void Crushers

I have shared one with the scientists where running wiggle at CI zero just moves a tiny bit.
(Wiggle at CI 0 almost does nothing)

beta_helix's picture
User offline. Last seen 5 days 18 hours ago. Offline
Joined: 05/09/2008
Groups: None

I am posting Round 2 for this puzzle WITHOUT the filters (so you will be able to let us know if that makes the difference).

You can help us with this by trying out this devprev puzzle and telling us if you run into the same issues as you had with this one:
http://fold.it/portal/node/992823#comment-23390

Thank you for your help and patience with this nasty bug!

Joined: 09/21/2011
Groups: Void Crushers

Same problem still exists on my shared with the scientists one from round 1. Actually it gains points at CI 0.

jflat06's picture
User offline. Last seen 1 day 3 hours ago. Offline
Joined: 09/29/2010
Groups: Window Group

As far as I can tell, it quickly gains some points, and then immediately after starts to crumple and have terrible score. It's definitely odd, as usually knocking out the Van der Waals score term completely messes everything up.

However, it is completely possible that the energy gradient of the CI-less score function matches that of the normal score function. This alignment of the energy landscapes wouldn't last long, which is why it crumples immediately after.

jflat06's picture
User offline. Last seen 1 day 3 hours ago. Offline
Joined: 09/29/2010
Groups: Window Group

All of the solutions that I have tried have resulted in the backbone crumpling and a terrible score. Is this still happening for people?

MurloW's picture
User offline. Last seen 8 weeks 1 day ago. Offline
Joined: 11/21/2012

I can wiggle 727 at 0.0CI with movement no greater than micropoints.
On the other hand, when I open other players' solutions (for evo) this doesn't happen, and they crumple as usual.
Also: wiggling at low CI (0.2 or whatever) doesn't compact it like on most puzzles, unless I disable backbone constraints during the low CI wiggle and quickly enable them again. That seems to loosen it up and give me literal wiggle room.

MurloW's picture
User offline. Last seen 8 weeks 1 day ago. Offline
Joined: 11/21/2012

I shared a solution to scientists to check it out, 727. Called: 0.0CI wiggle -- jflat06

jflat06's picture
User offline. Last seen 1 day 3 hours ago. Offline
Joined: 09/29/2010
Groups: Window Group

I've done some investigating, and as far as I can tell, this is the expected behavior of the minimizer in Rosetta.

The clashing importance slider modifies the fa_rep score weight in the score function that is used by the minimizer.

fa_rep is the repulsive score term that gives you a penalty when things are too close. In the solutions where this stickiness exists (that I've seen), there aren't atoms from different residues that are close to each other. In situations such as this, the penalty from fa_rep is minimal to begin with, and making it 0 doesn't affect the energy landscape very much. The other score terms (such as hydrogen bonding, bands, etc) are much stronger in this situation, and so they hold the protein in its current shape even when CI=0.

It might *feel* wrong, but there's no guarantee that the CI slider/minimizer should behave in an intuitive way.

MurloW's picture
User offline. Last seen 8 weeks 1 day ago. Offline
Joined: 11/21/2012

Okay, so...

still pretty weird then that when I open a shared solution by another player that evo'ed my solo, it behaves as expected (crumpling together with 0.1 - 0.3 CI wiggles);
and by the time I've evo'ed it (through mutation, for example) it won't budge by more than micropoints at 0.0 CI ? It seems like being in *my* client "infects" it with stickiness.. O.o

Keep in mind this happens on 727, like, all the time for me.

jflat06's picture
User offline. Last seen 1 day 3 hours ago. Offline
Joined: 09/29/2010
Groups: Window Group

These issues likely happen on 727 more often because the endpoints of the designable segment force it to stretch out, which prevents it from forming conformations which would collapse on themselves. The nature of your solutions are probably favorable in the fa_rep-devoid energy landscape, because you are optimizing for the other score terms.

MurloW's picture
User offline. Last seen 8 weeks 1 day ago. Offline
Joined: 11/21/2012

I see.. I think I kinda understand .. :p
Okay, thanks for all the help and info!

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