Restore the old rebuild settings

Case number:845813-987371
Topic:Game: Tools
Opened by:Mark-
Opened on:Friday, February 26, 2010 - 10:32
Last modified:Thursday, February 9, 2012 - 21:36

... or at least modify them. As it currently stands the new rebuild will try many options close to the starting point and very few, if any, far away from it. Useful if the current structure is close to where you want it, less than useless when it's not. As it stands, 'freestyle' and 'quest to the native' puzzles have just become much more difficult without the judicious use of bands and manual pulling of segments. Not something that will improve game play.


(Fri, 02/26/2010 - 10:32  |  21 comments)

Joined: 05/03/2009
Groups: Contenders

Absolutely couldnt agree more. You've changed rebuild into an end-game tool - at best, which is not overly helpful. Structure mods in freestyle used to snap into place when rebuilding, at least giving you a direction to work towards - now you've 'dulled the resolution'. Where before, I could run rebuild to 4 positions, and get, let's say, answers of 6,2,1 and 9 - I now get 500 positions of, then another 500 of etc etc. As a start-game strategy, particularly for freestyles, it's dead in this new form. As a tool to help, it's become redundant.
We play, we report back from the front-line.

Please set this as a high priority fix.
It's like waiting at a red light in traffic, and trying to judge if a 3 photon change in brightness, means I can get to work on time.


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

Okay, I'll look into fixing this.

Joined: 09/18/2009
Groups: SETI.Germany

Same here.
On freestyle puzzles, I often use rebuild almost at beginning to create helixes and to change the complete puzzle shape from "factory produced long rod" to a more handy and viewable clump.
And as rebuild mostly drops the score in the first moment (which is understandable and intended), it shouldn't be an endgame tool.

Thanks in advance,

Joined: 12/06/2008
Groups: Contenders

Rebuild used to snap a helix or a sheet into its proper shape, no matter how twisted I got it when I started folding a freestyle puzzle. Now... I can let the rebuild function go ten minutes, and not once see a curly helix or a straight sheet.

The new rebuild should be kept, but as an endgame tool. It does seem to make small adjustments that can help improve score. But please bring back the old rebuild tool, as well.

Joined: 11/09/2009

Just adding my voice to the chorus. I totally agree with all the points raised in this thread.

Thanks for looking into it, Seth.

steveB's picture
User offline. Last seen 11 weeks 2 hours ago. Offline
Joined: 01/08/2009
Groups: Void Crushers

Having given it a few days to try the new rebuild I am in agreement with everyone above. The new tool simply will not rebuild a helix to save its life, and does not 'snap' sections into place at all. Reverting to the old rebuild looks like a sensible option to me also.

Brick's picture
User offline. Last seen 1 year 2 weeks ago. Offline
Joined: 07/15/2008
Groups: Beta Folders

Agreed. Can't rebuild a helix at all. I left one for an hour, and it didn't look any different after all that wasted time.

PLEASE bring back the old Rebuild.

Joined: 05/09/2008
Groups: Void Crushers

Add my vote for the old rebuild. All my beginning puzzles scripts are pretty much reduced to end game use.
Please hurry :)

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

I've just posted an update that should restore the original rebuild behavior. Please reopen if there is still a problem.

Joined: 05/03/2009
Groups: Contenders
Status: Closed » Open
Type: Suggestion » Question

Odd to re-open a year old feedback, but 410, and the last few puzzles seem to be acting in this way again. Rebuilding is a black art indeed, but it just doesnt deliver now - either by hand or scripp

Joined: 05/09/2008
Groups: None
Type: Question » Bug

Thank you very much for reviving this feedback, CFC, I was actually thinking about this very problem this past weekend!
Hopefully we can investigate what is different between the original rebuild and the current one.

If you have any specific examples that could help us pinpoint any differences (because in March 2010 we did revert it to the original rebuild behavior, so it's odd that it is acting differently) that would be useful!

Joined: 06/17/2010

When there is selected helix and loop loop is rebuilded many times and helix is untouched.
Also setting band to rebuilded area in most cases not helping at all - tried rebuild sheet to fit other one, made several bands and rebuild tool not made it like it should loooong time... rebuilds in QTTN are also difficut.

IMO tool should "watch" bands and trying to make "proper" look of secondary structure corresponding to bands set.

Tool also sometimes only "flickering" and nothing is done in undo graph. Sometimes only for 1-2 segments, always when rebuilding part of sheet or helix.

Mark-'s picture
User offline. Last seen 2 weeks 1 day ago. Offline
Joined: 08/05/2009
Groups: Contenders

I have been wondering about this behaviour for a while now but just can't pinpoint exactly what has changed. It used to be relatively easy to rebuild small areas and wiggle them out to get rid of the 'red bits' but now the red often remains until the puzzle closes. Also, it seems the rebuild has to be exactly right otherwise it just wont wiggle out or it will pull adjacent areas out with it.

Assuming that wiggle and rebuild are a function of the backbone/sidechain score, maybe something in the scoring has changed in recent months? It definately seems to be more sensitive to something.

Joined: 06/17/2010

A bump.
Looks like it is somehow related to wiggle problems we have?

Joined: 05/03/2009
Groups: Contenders

Not sure rav...

Doubt it is wholly sourced in wiggle. Has been an issue for a long time. May even go back to the notion of every puzzle having individual wig and shake 'settings', so each one feels different. But, the last 3 have definitely felt like there's something different with the scoring function, in that it's only showing 'highlights' - it's the best way i can describe it. Where previously, you could gauge what was happening by the speed of change in the score display, ie the speed of it slowing down, when you wiggle, or the 'snap-snap-snap' when it shakes, no longer applies. It seems like you're only getting an 'edited highlight' of what occurred to move the score. If it's a change in score calculation, it might explain why shake is now sticky, wig is now lazy, and thus rebuild as a consequence of those 2, is a lot more bland and 'blunt'. Very much a WIP, but we can only report back from the front line. :)


spmm's picture
User offline. Last seen 13 hours 3 min ago. Offline
Joined: 08/05/2010
Groups: Void Crushers

I agree there is something different about rebuilds, in my limited experience I am only speaking about the last few months.
I used to band sidechains to loop positions, and then rebuild quite small areas, this was usually very effective. It just seems virtually impossible now, rebuild just flickers about and then takes a really strange position. I know the banding is correct and I'm not trying to force an unnatural position as I'm using a guide as a solution. Also not just on one puzzle. Helixes and sheets also don't seem to take good poses.
My two bobs worth.

Joined: 05/09/2008
Groups: None

spmm, you might be helpful in diagnosing when this change occurred since you started playing in August 2010, right? Do you have any idea around when you noticed this difference?

This is on our long list of things to fix, but if we could pinpoint where this particular problem started (at least to be able to diagnose the differences between now and back in the good old days) it would help us address the problem sooner.

Joined: 08/24/2010

The rebuild tool perhaps can be optimized in this way: In order to eliminate the tool returning similar conformation results that score differently, yet retain the same shape, (perhaps differing by only a small atomic difference, or mild reparameterization) - it may beneficial to map all near-solutions onto a curve C: Curves that are within a rotation and/or a re-parameterization of each other often result in different elements of C despite having the same shape. Let SO(3) be the predefined group of 3×3 rotation matrices and Gamma be the group of all re-parameterizations (actually positive diffeomorphisms of the interval [0,1]) For a curve, Beta, a rotation O, within the set of SO(n), and a re-parameterization, gamma 1, within the set of Gamma, the transformed curve is given by O(Beta dot gamma 1). The square root velocity function (SRVF), or rate of conformational change of the transformed curve, is given by square root(d gamma 1 / dt) * O ( q dot gamma 1 ). In order to unify all elements in C that denote the same shape, we should then define equivalence classes, Q. Each such class Q is associated with a shape uniquely and vice versa. The set of all these equivalence classes is called the shape space S. Mathematically, it is a quotient space of the preshape space: S = C/(SO(n) cross Gamma)

"A Mathematical Framework for Protein Structure Comparison" -

Joined: 06/17/2010

Or just allow us pull/freeze any atom :)

Joined: 05/09/2008
Groups: None
Status: Open » Open

we are looking into this again, but any help or feedback about this would be appreciated!

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

I've managed to get the very original version of rebuild working in the game again. It'll go out in a future release.


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