Back to Recipes Homepage
recipe picture
Recipe: Loop rebuild 8.0
Created by spvincent 66 56
5
Your rating: None Average: 5 (2 votes)

Profile

Name: Loop rebuild 8.0
ID: 102235
Created on: Fri, 11/04/2016 - 18:36
Updated on: Fri, 11/04/2016 - 18:36
Description:

See first comment



Best For


Comments

spvincent's picture
User offline. Last seen 3 hours 34 min ago. Offline
Joined: 12/07/2007
Groups: Contenders
Principle change from 7.0 is

Principle change from 7.0 is the addition of remix, which is enabled by default. When using remix it'll try to do the same number of remixes as specified in the N rebuilds slider (default 8), then shake each one and wiggle the two best in a similar manner as for rebuild (see Loop rebuild 7 doc). If it can't find any rebuilds there'll be a message in the output window "Skipped".

Other changes, mostly minor, are that bands are kept by default, there's a new filters-only scoring mode, and an options failed box to reduce visual clutter.

Loop rebuild 7 documentation below:

Not too many changes since 6.0 to really warrant a full new version number. But never mind: integers are cheap and there's no danger of running out.

Changes since 6.0

a) More varied secondary structure selections
b) On improvement, does an idealize on the rebuilt segment.
c) Prints out mutation data and limited performance data on termination.
d) General tidying up.

----------

Given a number of rebuilds, which is the best one to use? Not necessarily the best-scoring one, or even the best-scoring one at low CI (see http://proteinoftheday.wordpress.com ), but probably the best-scoring after shake and, which is what's going on here. Since Shake and Wiggle are computationally expensive, some effort is made to work only with the relevant sub-section of the protein.

Can be used at any Wiggle Power.

Algorithm outline


For each section to rebuild
   For each rebuild (defined by N Rebuilds)
      Rebuild once
      Set selection to that defined by Inner sphere
      Shake
   end
   
   For each pose corresponding to the two best scores above
      Set selection to that defined by Outer sphere
      Wiggle
   end

   Take the best pose from the two above
   Set selection to that defined by Outer sphere
   Shake (or Mutate if a design puzzle and that option chosen)
   Wiggle again if the Shake or Mutate gained anything.

   If we've gained an overall improvement
      Idealize and Wiggle
      Record overall improvement
  end

end

Dialog settings

Rebuild Length

How big a section to rebuild. Values from 4 to 6 are typically most useful, but larger values are valuable earlier in the game.

Min residue, Max residue

Only rebuild sections that lie within this range: by default the entire protein can be used. These values can be usefully changed to whatever region of the protein you're working on: it's often desirable to set them to the start and end of a loop you're working on.

Rebuild criterion

Slider with 4 values that controls, within the range defined above, which sections will get rebuilt based on their secondary structure.

1) Loops only
2) Loops allowing one non-loop residue.
3) Loops or sheets
4) Anything

Convert to Loops

(Not meaningful if Loops only above is selected). If checked, non-loops will be converted to loops before rebuilding: if not the secondary structure will be maintained.

So for example, if you had a helix from 35-40 and you wished to rebuild it as a helix with, say , a loop spacer 2 residues wide on either end, suitable settings would be:

Rebuild Length 10
Min residue 33
Max residue 42
Loops only unchecked
Convert to loops unchecked

N rebuilds

How many rebuilds to attempt. The default is a bit of a compromise: too small a value means that not enough poses will be created: too many and there's a high probability of ending up working on a pose which is extremely similar to the starting structure.

Mutate not shake

This checkbox will appear if and only if the puzzle contains mutable residues. If set (default), the sidechain shake in the algorithm will be replaced by a mutate.

Post wiggle threshold

An optimization used to prevent doing unnecessary shake/mutates. If the score after the best rebuild/shake (as described in the algorithm) is below the best by more than this amount, then skip the subsequent fuze-like processing as an overall gain is unlikely,

The default value works quite well: it might be worth lowering it to say 10 for endgame running or raising it to 60 or so at the beginning, particular for symmetry puzzles where overall scores are greater than usual.

Score type

Allows different scoring functions to be used - the values are:

1) Normal. Self-explanatory: the value as described by the program. Most of the time this is what you'll want

2) Exploration score. Perhaps one day there will be another exploration puzzle.

3) Energy score. Design puzzles without those filters.

4) Hiding. Just use the hiding component of the score.

5) Bonding. Just use the bonding component of the score.

In the last two cases, no fuzing will be done, irrespective of the threshold value.

6) As explained above: Toggles Slow Filters apart from scoring.

Check conditions met:

If set, check that conditions are met before accepting any improvement. For puzzles with RMSD conditions, etc.

Joined: 09/24/2012
Groups: Go Science
suggestion

Could you find a way to let it run on "no rebuild" puzzles (with only remix) ?

spvincent's picture
User offline. Last seen 3 hours 34 min ago. Offline
Joined: 12/07/2007
Groups: Contenders
The "Use remix" checkbox

The "Use remix" checkbox should be enabled irrespective of the kind of puzzle.

One thing I've found: it can be helpful with Remix to use the :Convert to Loops option.

Joined: 09/24/2012
Groups: Go Science
Suggestion

Remix replaces rebuild here. May be an option to try rebuild in case there is no possible remix (optional because of no rebuild puzzles).

Want to try?
Add to Cookbook!
To download recipes to your cookbook, you need to have the game client running.
Parent

none

Children
Authors
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, Microsoft, Adobe, RosettaCommons