A rebuilt sheet should be... a straight zigzag

Case number:845813-993724
Topic:Game: Tools
Opened by:Bletchley Park
Status:Open
Type:Suggestion
Opened on:Friday, October 12, 2012 - 21:26
Last modified:Sunday, October 21, 2012 - 18:01

Whenever I rebuild sheets, I have to band the things up, be very patient or just say a prayer that it ends up as a straight zig-zagged structure.

Can rebuilding sheets please by default produce a zig-zag straight structure so we can spend time on doing other things ? As a straight zig-zag structure is what we want with a sheet.

(Fri, 10/12/2012 - 21:26  |  18 comments)


brow42's picture
User offline. Last seen 1 day 21 hours ago. Offline
Joined: 09/19/2011
Groups: None

Rebuilding in structure mode is nice...but it rarely works. 99.9% of the trial positions are rejected. I always, always end up just changing it to loops.

It's TRYING to make a zig zag, but it just can't make the ends meet. Need to use cutpoints to make it work.

Joined: 05/19/2009
Groups: Contenders

I think the algorithm to figure out what is important for the structure needs a change. Even with plain-obvious free space, an open end and nothing to obscure a straight path the algorithm very very often comes up with bent sheets. This is a waste of time, because we all know we want that zigzag straight line.

Maybe the segments that have sheet property and are connected should have absolute priority over any loop arrangement connected to it unless the loop is frozen.

Generally it works fine for helices, why not for sheets ?

marie_s's picture
User offline. Last seen 1 year 49 weeks ago. Offline
Joined: 05/18/2008
Groups: None

The tweak tool, unfold the sheet on the same manner on both side beginning with the second segment, it will be better for me if one side unfold from the beginning of the sheet and one side from the end. Which one wich? arbitrary , we will try.
In the meanttime, the quicker way to have a straight sheet is to put a band at each end in the opposite direction and wiggle to stretch

spvincent's picture
User offline. Last seen 3 hours 41 min ago. Offline
Joined: 12/07/2007
Groups: Contenders

I'm not sure it's really desirable that sheets should come out as a straight zig-zag, geometrically pleasing though this may be. The sheets in real world proteins such as those in the RCSB protein data bank always seem to be bent or twisted in some fashion.

Joined: 05/19/2009
Groups: Contenders

They will end up being twisted or bent eventually through scripting based on better scoring but that is not how I would like them to be when starting my folding. In most cases I have seen, having the straight zig zag is a desireable shape. Maybe make it a switchable property ?

Joined: 05/28/2012

I agree with marie_s' approach, although i just use tweak. all you need to do is cut-point the sheet off and straighten it out. no rebuilding required. Same thing with a helix by the way. rebuild is essentially there to randomize a rebuild, not to necessarily make it better.

Joined: 05/19/2009
Groups: Contenders

"Rebuilding in structure mode is nice...but it rarely works." supports my point.

"The tweak tool, unfold the sheet on the same manner " It unfortunately does not unfold misformed sheet-like structures (zig-zig-zag) into a proper sheet (zag-zig-zag). I have verified this to be true.

"I'm not sure it's really desirable that sheets should come out as a straight zig-zag" I would like to have that as a starting point. I can then decide to bend or misform it.

"all you need to do is cut-point the sheet off and straighten it out. no rebuilding required. Same thing with a helix by the way. rebuild is essentially there to randomize a rebuild, not to necessarily make it better." You'll be amazed what you can learn here.

spmm's picture
User offline. Last seen 40 weeks 5 days ago. Offline
Joined: 08/05/2010
Groups: Void Crushers

also I can live without wriggling out cut points - bores me silly

Susume's picture
User offline. Last seen 1 week 2 days ago. Offline
Joined: 10/02/2011

I don't think it would be desirable for rebuild to default to straight sheets when running in a script. I often run a rebuilder with structures on in early game to help keep the general shape of the protein, and those sheets need to be able to twist and curl a reasonable amount while still remaining "sheety." I agree with using the cut tool and tweak, or wiggle or rebuild with bands at both ends, to create a completely straight sheet.

Joined: 05/19/2009
Groups: Contenders

You have a good point with your script remark, but the part where you mention the other means to create a ZIG-ZAG-ZIG sheet from a ZAG-ZAG-ZIG sheet is unfounded.

Please show me an example where such a configuration (ZAG-ZAG-ZIG) was turned into ZIG-ZAG-ZIG without using rebuild and without extraordinary effort in movements and time, I'd like to learn how to do that.

Susume's picture
User offline. Last seen 1 week 2 days ago. Offline
Joined: 10/02/2011

Now I understand the problem better. I used the cut tool to create a sheet with sidechains facing A-B-B-A. Tweak was able to get the sidechains back on the right sides, but the backbone then had a bent zag-zag shape. I was unable to straighten it using any of the methods mentioned above (tweak, rebuild with and without bands, wiggle with strength 10 bands on ends and/or sidechains). The only way I found to fix it was to cut again between the two offending segments and rotate half the sheet with the move tool (which was how I created the problem in the first place). After rejoining, tweak alone was still unable to make it nicely straight; that required some rebuild followed by tweak.

I was surprised that rebuild could not find a good solution, given that the sheet was cut out and completely free to move (no endpoints to match up to). Either tweak or rebuild needs to be changed to recognize a zag-zag problem and correct it.

Joined: 05/19/2009
Groups: Contenders

I applaud you for your explanation and effort Susume ! Thank you.

Joined: 05/28/2012

So if the only way you get to a ABBA in the first place is if you use cut-points, then the solution would be to just not put it like that. It Makes sense because ABBA is completely unnatural. It would be like cutting your arm off mid way flipping it around putting it back on then expecting it to still bend and function the right way. Your not gonna have a good time.

Joined: 05/19/2009
Groups: Contenders

You miss the point. Susume did that to test the hypothesis that you could, as you incorrectly stated, use tweak to get any shape into a straight zig-zag. She proved that that does not (always) work. In some of the models we work with, a zag-zag-zig structure may appear, it is then needed to run rebuild to get it right. My point is that rebuild should always search for a zig-zag-zig solution rather than present us with less useful shapes which may still contain zag-zag-zig and may still be bent to a useless point. If you have to run rebuild anyway, you may as well go for the straight shape.

I posted a suggestion. It is up to the devs to see if that makes sense and put it into practice or not.

brow42's picture
User offline. Last seen 1 day 21 hours ago. Offline
Joined: 09/19/2011
Groups: None

Ah! I wish I had come back to this thread sooner. A picture would have been great! You're talking about flipped segments in sheets? (cis bonds) I've thought a lot about these.

irc_269484_1350496864.png

I'm surprised you're having trouble with this...they occur with very low probability and disappear in the very next rebuild. They're very hard to make on purpose. I've literally spent days trying to hand-flip glycines with rebuild in Snow Flea. My Glycine Flipper script is a complete failure. I vote no to eliminating flips from rebuild completely.

Tweak absolutely does not affect cis bonds in any way, only rebuild. And, that twist doesn't just vanish, it has to go somewhere. It might move to a u-turn, or it might turn into a twist along the sheet, ruining it.

These flips do happen in nature and usually for a good reason. The picture above is from a CASP refinement, so presumably is close to the native. Why would we want a flip? If it's in a sheet, it's probably part of a beta bulge.

A beta bulge forces the sheet to bend (not in the thin bendy direction, but in the thick direction). We've seen these a lot in puzzles where the sheet has to wrap around something, but not like in a barrel. If you have a proline in the middle of a sheets, it's probably there for this express purpose (prolines are 3x as likely to have a flip).

You have a nice functional protein, but then there's a mutation and you have an extra segment. If you just have a flat sheet, everything after the insertion is now upside down, with your 'phobes and 'phillics swapped. Better to just include an extra twist.

A flip might be the best way to bury a big 'phobe or unbury a big 'phillic.

Certain xxx-pro beta hairpins (a tight turn connecting to sheets) ALWAYS have a flip. Without it, a sheet would be upside-down!

More on rebuilding in structure mode: I say it doesn't work because it just SITS there. I don't get any rebuilds at all. I either have to use the scissors tool and rebuild it while free floating, or I have to rebuild it as loop at least once, then set it back to sheet and rebuild the (bent) sheet. It's very frustrating to have 100% of rebuilds rejected in structure mode.

As an aside, banding also makes the rejection rate go WAY up...and the only way to get a cis bond with rebuild on purpose is...banding.

I can't preview this post so I hope it all turns out.

brow42's picture
User offline. Last seen 1 day 21 hours ago. Offline
Joined: 09/19/2011
Groups: None

My picture is missing so here it is:

Susume's picture
User offline. Last seen 1 week 2 days ago. Offline
Joined: 10/02/2011

Thanks for the info brow! My picture is different from yours, but I think they're two forms of the same thing. Tweak put the sidechains back to alternating sides (ABAB) but now there are 2 right turns in a row, at either end of the blue asparagine. I guess that matches your description of a beta bulge. I ran rebuild on this for about 10 minutes, which found roughly 350 positions. None of the poses had the clear right-turn-left-turn shape I was looking for. Most had two right turns in a row. About 5% were straight between the asparagine and tyrosine so I couldn't tell for sure if they had switched to right-left or not. Four poses had a twist in them (like your picture) and the sidechains back on sides ABBA. Maybe I'd have gotten a different assortment if I had a different sequence of aas. Seems like there are two sides to this problem: when you have a zag-zag you don't want, it's hard to get rid of, and when you want to add one (or move one along the backbone I guess), it's hard to create.

Interesting to learn that these things occur in nature and are sometimes needed in a protein. I'd still like a way to get rid of them though.

zag-zag (two right turns) in a sheetzag-zag (two right turns) in a sheet

jeff101's picture
User offline. Last seen 5 hours 56 min ago. Offline
Joined: 04/20/2012
Groups: Go Science

I recently posted a recipe called makehelix (http://fold.it/portal/recipe/42111) that uses bands to make a variety of helix types. It can also make ideal beta-sheets. Perhaps it can solve some of your problems.

If you think of ways to improve makehelix, please e-mail me your suggestions or adapt the code as you see fit. Thanks!

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