Case number:845813-2000783
Topic:Game: Tools
Opened by:Enzyme
Opened on:Monday, June 1, 2015 - 01:50
Last modified:Wednesday, August 16, 2017 - 22:12

Remix should be available on all puzzles.
For years the best tool we have had to create beta hairpins has been Rebuild which can take over an hour of testing different positions before rebuild will find one acceptable hairpin.
The longest it has taken me to correct a hairpin with remix is 20 minutes.
After finally giving us tools to help cope with the disfunctional rebuild, it makes no sense for one of them to be unavailable on any puzzle

(Mon, 06/01/2015 - 01:50  |  6 comments)

jflat06's picture
User offline. Last seen 23 hours 24 min ago. Offline
Joined: 09/29/2010
Groups: Window Group

We don't provide Remix on non-design puzzles because it operates on a reduced fragment set, which is bad for prediction puzzles.

The loops that Remix gives you are 'ideal for design'. They are common and stable fragments that have a good chance of folding up, and you can design your protein to incorporate them.

However, in prediction puzzles, the actual loop or fragment that belongs in a certain area of the protein might not even be available from the Remix tool. Thus, if you were to use Remix, you would end up with nothing but incorrect fragments.

Joined: 07/10/2008
Groups: None

If Rebuild cannot find correct structures and Remix cannot provide correct structures. Looks like we have a Mexican standoff

Joined: 07/10/2008
Groups: None
Susume's picture
User is online Online
Joined: 10/02/2011

I'm guessing from your screenshot that you are not satisfied with the shape of the beta hairpin turns. Two ways to improve them are to make sure they are made with appropriate amino acids (not every sequence of aminos can make a nice 2-residue turn) and to fan the strands (like a chinese fan) so the sheet is not a straight wall. I use this link as a guide in making 2-residue hairpins: (You may already know this stuff, but someone else may find it helpful.)

The shape of hairpin you are getting is Type I, where both aminos adopt an alpha helix shape. Looking at the turns end-on as in your picture, this shape pushes the right-hand strand end down relative to the left-hand, which is the opposite of what you want for the more common type I' or type II' turns. Type I' and II' turns require one or both residues to be glycine (see link) so if you don't have that, the type I helix-like turn may be the best you can get. To make type I' or II' possible, make sure your aminos are gly+gly, asp+gly, asn+gly, or gly+polar.

To give the turns better end-points for forming a type I' or II' turn, cut the strands apart and try fanning them (rotating from the center point) so each left hand strand end (facing you as in the pic) is a little lower than the one to its right. Rotating the strands from their midpoints will also place the farther ends of the strands in the same relationship. Strand ends 2 and 3 as highlighted in your pic are in the desired relationship, so you can just rotate strands 1 (so its near end is lower) and 4 (near end higher) to get them all in a fan.

This image shows the desired fan shape:

I can't promise this will make rebuild give you pretty 2-residue hairpins, but it will improve your odds.

Joined: 07/10/2008
Groups: None

One of the 2 segment hairpins is gly-tyr. The other is asp-gly. And I know rebuilding for hours can give good hairpins. My problem is the sheer amount of time wasted, to quote your link, "one of the simplest supersecondary structures". Many puzzles it is not as great of an issue. The two puzzles where it bothers me the most are the last two de-novo puzzles where literally half of the build time needed to create a fold is spent trying to keep the hairpins together

Joined: 07/10/2008
Groups: None
Status: Open » Closed

fixed at some point, but not as a result of player feedback.


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