Questions on Remix tool

Case number:845813-2002995
Topic:Game: Tools
Opened by:gitwut
Status:Open
Type:Question
Opened on:Tuesday, October 25, 2016 - 15:34
Last modified:Tuesday, November 15, 2016 - 16:43

I'm using Puzzle 1299 as an example and am using client: 20161006-39881adc28-win_x86.

1299_1.png shows the initial position. The remix tool gives 4 ideal loop suggestions that raise questions about how the tool works.

What's odd is that the score for the 4 ideal loops are all lower than the score for the bad loop. Why is that? Is it because of the resulting poor ideality, backbone scores for segment 25? How is it that these can be ideal loops with such bad backbone scores (particularly the bad angles)? See Images 1299_2.png through 1299_5.png for examples.

AttachmentSize
1299_1.png227.71 KB
1299_2.png201.53 KB
1299_3.png200.12 KB
1299_4.png210.65 KB
1299_5.png211.06 KB
(Tue, 10/25/2016 - 15:34  |  6 comments)


gitwut's picture
User offline. Last seen 33 weeks 3 days ago. Offline
Joined: 05/18/2012
Groups: Contenders

I would appreciate feedback on this from a dev. If it is unexpected behavior, fixing it would make the tool much more useful. If expected, knowing why would be useful as well.

jflat06's picture
User offline. Last seen 19 hours 17 min ago. Offline
Joined: 09/29/2010
Groups: Window Group

It's a little hard to tell just from the screenshots given. The Backbone score is composed of several parts -

Rama, omega, p_aa_pp, angle/atom_pair/coordinate/dihedral/dof constraints.

My guess of what is happening here:

Rama and p_aa_pp are both give evaluations of how well your backbone angles match the sequence that has been assigned. Remix doesn't take your sequence into account when it selects a shape, so it's actually fairly likely that your new shapes will score lower on backbone *until* they are redesigned.

The shape is good, it just might not match the sequence that you already had there.

Joined: 09/24/2012
Groups: Go Science

"redesigbed": do you mean mutated to other AAs (with rama map open to verify the angles)?

jflat06's picture
User offline. Last seen 19 hours 17 min ago. Offline
Joined: 09/29/2010
Groups: Window Group

Yes, you will need to change your AAs after remixing for the rama maps to be correct (usually).

Susume's picture
User offline. Last seen 2 days 21 hours ago. Offline
Joined: 10/02/2011

Interesting, that suggests we should use mutate in any script that does remix, because the remixed poses might not score well until the AAs are changed.

jflat06's picture
User offline. Last seen 19 hours 17 min ago. Offline
Joined: 09/29/2010
Groups: Window Group

That sounds correct. I believe this is one of the reasons we didn't used to give you remix in prediction puzzles. We ended up doing it anyway, because Rebuild really wasn't any better.

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