Feature request: ability to fold proteins as several smaller pieces

Case number:845813-986943
Topic:Game: Tools
Opened by:infjamc
Status:Closed
Type:Suggestion
Opened on:Friday, October 23, 2009 - 01:25
Last modified:Tuesday, February 7, 2012 - 23:10

In certain puzzles, cutpoints are added to specific locations as constraints, to ensure that two neighboring residues have the proper distance between them. Now, given that freestyle folding puzzles start with an extended configuration, it would be really helpful if we could add our own cutpoints so that smaller segments of the protein could be manipulated separately.

==> Why is this necessary? Currently, there is no way to change the backbone configuration without affecting the rest of the protein unless one freezes off certain segments. But if segments are frozen, the endpoints would be fixed, and thus certain manipulations would be impossible (e.g. straightening a helix). The solution? Cut the protein into pieces, fold the smaller pieces separately, and then glue them back together. Of course, the cutpoints need to suffer from a backbone score penalty modified by distance, and another behavior tab for backbone score would be needed so that the pieces would not come back together in undesired ways until the folder wants to glue them back.

(Fri, 10/23/2009 - 01:25  |  13 comments)


Joined: 12/06/2008
Groups: Contenders

Here, here! Anyone else frustrated trying to rebuild a helix that's in the middle of a freestyle puzzle?

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

The idea of changing small separate pieces is good, but if my old computer can handle it?

Brick's picture
User offline. Last seen 2 years 19 weeks ago. Offline
Joined: 07/15/2008
Groups: Beta Folders

Sounds like a great enhancement to me!

I generally give up on the freestyle puzzles because of this issue, and just let them roll up into the best scoring yarn-ball I can come up with.

Brick's picture
User offline. Last seen 2 years 19 weeks ago. Offline
Joined: 07/15/2008
Groups: Beta Folders

Sounds like a great enhancement to me!

I generally give up on the freestyle puzzles because of this issue, and just let them roll up into the best scoring yarn-ball I can come up with.

Brick's picture
User offline. Last seen 2 years 19 weeks ago. Offline
Joined: 07/15/2008
Groups: Beta Folders

Sounds like a great enhancement to me!

I generally give up on the freestyle puzzles because of this issue, and just let them roll up into the best scoring yarn-ball I can come up with.

infjamc's picture
User offline. Last seen 12 hours 4 min ago. Offline
Joined: 02/20/2009
Groups: Contenders

Brick:

Uh, you've just triple-posted. ;-)
As for the yarn ball-- sadly, that is my current structure in 212. It does score higher than a giant helix...

Crashguard:

As far as I know, proteins with cut-points in them should not be a problem for older computers if they are already able to handle the recent large puzzles (i.e. those with >150 protein segments with ligands or DNA). After all, this change should only affect the continuity between residues.

For example, if residue #14 is connected to residue #15 (which is the default case), then changing the configuration (e.g. phi and psi angles-- see http://en.wikipedia.org/wiki/Dihedral_angle ) of residue #14 would move everything downstream unless the endpoints are fixed. If a cut-point is inserted between #14 and #15, however, then changing the figuration of #14 would hold #15 in place while creating a gap between the two residues. While the gap is undesirable, this way of manipulation small segments has the advantage of keeping the configuration of everything downstream relatively intact (assuming that a subsequent wiggle is sufficient to close the gap)-- which has many practical uses.

==> Bottom line: the technical complexity of representing proteins with cut points should not be higher than existing puzzles with multiple pieces (whether they might be proteins, DNA, or other ligands).

***FoldIt developers: Please correct my assertions above if they're wrong-- I'm writing this based on my limited knowledge of how Rosetta works (cf. fold trees, which ensures that changes to the configuration of a residue would not propagate outside a small area). Now, I do realize that implementing the ability to manually do loop modeling (which would involve defining the endpoints of fold trees and location of cut point *plus* a change in the score function) can take quite a while, and that I only know the technical details because I had worked with Rosetta 3.0 in the past (i.e. this is probably not public knowledge even to those who have the background in biochemistry).

...Hence, I'll drop the priority of this project down to a 5 for the time being. (After all, the scores from freestyle puzzles seem to indicate that some of the top folders are able to reconstruct entire proteins from scratch even without the ability to cut them into smaller pieces-- so I guess the problem is me rather than the program?)

Brick's picture
User offline. Last seen 2 years 19 weeks ago. Offline
Joined: 07/15/2008
Groups: Beta Folders

Sorry about the triple posting. The web site just returned a blank white page after I clicked "Save". I just did a reload a couple of times until a page appeared (with 3 posts) ;(

beta_helix's picture
User offline. Last seen 17 hours 24 min ago. Offline
Joined: 05/09/2008
Groups: None
Status: Open » Open

great idea that we have talked about implementing many times...
sadly, other things always end up getting higher on the priority todo list.

Joined: 06/17/2010

Look there: http://fold.it/portal/node/990194
It is (was?) implemented in standalone client.

infjamc's picture
User offline. Last seen 12 hours 4 min ago. Offline
Joined: 02/20/2009
Groups: Contenders

If you're referring to the trim function... I would say "close, but not quite."

Here's why: When a protein in the standalone client is trimmed, the ends are fixed to prevent the hidden parts from being affected. This is still not quite the same as literally cutting the protein into two or more pieces that can be moved independent of each other.

Joined: 06/17/2010

Yes, but if you do so distances and positions between start and end segemtns of edited part can be changed. Then overall structure will collapse when merged. Or it need use some ugly strech&rebuild to fit pieces back together - score will drop rapidly.
Trim would be enough. Just trim another part (longer) and work on it. Need focus? Trim only few segments and wiggle them to death.

infjamc's picture
User offline. Last seen 12 hours 4 min ago. Offline
Joined: 02/20/2009
Groups: Contenders

Good point. However... I vaguely remember that Rosetta has several protocols, such as something called CCD ("cyclic coordinate descent"), that could be used to close the gaps created by the usage of cut-points.

For the purpose of foldit implementation, I would imagine that a method similar to what's being used by the alignment tool for filling in gaps could be used-- namely, it would try to close the gap via a loop rebuild if possible, and tell you that the gap is too wide to be closable if it fails.

infjamc's picture
User offline. Last seen 12 hours 4 min ago. Offline
Joined: 02/20/2009
Groups: Contenders
Status: Open » Closed

Closing old case since the alignment tool in the developer preview now uses cutpoints.

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, Boehringer Ingelheim, RosettaCommons