Back to Recipes Homepage
recipe picture
Recipe: Ebola Rebuild 1.12.6
Your rating: None Average: 4.5 (47 votes)


Name: Ebola Rebuild 1.12.6
ID: 48385
Created on: Sat, 03/01/2014 - 14:06
Updated on: Tue, 12/01/2015 - 19:34

DRW+GAB, NC ready. By Jean-Bob and L'Alliance Francophone.

Best For


Joined: 09/24/2012
Groups: Go Science
Best for middle to end game

This recipe combines the power of the rebuilds from DRW with the ability of GAB to increase the score by banding some segments.

Moreover, we made some improvements :
Whole prot : The possibility to rebuild and GAB one by one each segments of the protein, from segment 1 to segment max.
ScoreFilter : Select the minimum score required to launch GAB. Tick "Do not launch GAB" if you only want to rebuild and fuse.
Allow loss : Ebola will not restore the highscore if the score is included in this Loss threshold
FastFuse : Use 1 fuse instead of 5
Saving module : If you tick the "Start again since the last stop" box, Ebola will restart at the same segment as before stopping the last Ebola run. The data from the last run is stored in a note on segment 1
Language : The structure of the script permits to translate very easily the dialogs in many languages.

"More" options :
Rebuild Filter : Skip wiggle after the rebuild if the score is not included between these 2 values compared to the highscore (highscore-Minfilter Subscore management : Consider only one of these subscores to define the worse segments
ForceZone : Rebuild only a selected part of the protein. It is possible to enter many zones. Tick "Walk in the Forcezone" to rebuild from the first to the last segment each zones.
Launch automatically GAB : Launch automatically GAB if the threshold is exceeded
Precise band : Bands from GAB always start from next to the rebuild segments.
Display more information : Display more information in the output of the recipe

"GAB" options :
Number of critters, generations and allowed loss

Nevertheless, it takes a lot of time for each cycle. Excellent recipe for an overnight folding!

ver 1.12.4: Fixed bugs and NC optimized

Note: for sepsis and hotspots, only "whole protein" will work.
Sorry for that. Contextual options will only be implemented later on.

Joined: 09/24/2012
Groups: Go Science
Wiggle power

Best with Low Wiggle power to Medium wiggle power

Joined: 09/24/2012
Groups: Go Science
Ebola Rebuild options (new labels)

Default: Select zones of Segments min length 3 to Segments max length 6.
For each length, Try worst 10 zones or Whole protein.
For each zone (worst or all), test 15 Rebuild/zone: rebuild + shake or Mutate instead of shake. Here, if the loss is not too big (see second dialog panel), wiggle selected unless Skip Fuse (much faster) is checked.
Of these 15 Rebuild/zone, keep 3 best Solutions/zone and fuse them 5 times or Fast Fuse (faster, less points) 1 time or even Skip Fuse (much faster) if you want to preserve the structure of your protein on early begin game.

If the loss then is less than GAB Filter (default: 20), launch GAB or Do not launch GAB.

Before all of this, you may Transform protein to loop or keep secondary structure (like on early game in order to optimize your design).

Main menu Ebolarebuild1124mainEN.PNG
Segments min: number of segments for smaller zones to start with
Segments max: number of segments for bigger zones to end with (then it starts again with small zones)
Try worst: Takes the x worst zones of this length
Whole protein: Takes all possible zones of this length
Rebuild/zone: For each zone, rebuild(1)+(shake or mute)+(wiggle selected)
Solutions/zone Keeps this number of best solutions before to fuze and/or GAB
-------------Threshold for GAB---------
GAB Filter: allowed loss before to launch GAB
Do not launch GAB: never GAB = DRW only
Allow loss: Keeps the best zone of a specific length, even if it scores less than the best score!
loss threshold: There is a limit of course (otherwise, the recipe will ends to loose only)!
Transform protein to loop: it gives more freedom to rebuilds, for end game
Skip Fuse (much faster): good to keep the structure, on begin game
Fast Fuse (faster, less points): 1 fuze instead of 5
Mutate instead of shake: (just after each rebuild), available only on design puzzles
Continue with former options: If Ebola has already been used on this protein, former options are most of the time saved in note from seg 1. Ebola will start again with the same options where it stopped if options are present in note.
Buttons: OK, Cancel, More, GAB

Joined: 09/24/2012
Groups: Go Science
The picture of first dialog box:


Joined: 09/24/2012
Groups: Go Science
oops duplicate


Joined: 09/24/2012
Groups: Go Science
Advanced options

When pushing on "more" button, you get this:
This dialog deals with what happens on the rebuild rounds.

----------Rebuild filter--------------
Min filter: Above this loss, after rebuild and shake, the solution will not be wiggled
Max filter: Skip Fuze and GAB if rebuild+shake+wiggle does not move at least by this amount

----------Subscore management--------------
Score types:Score for selecting worst scoring zones.
1= All
2= Backbone: clashing + packing + backbone + ideality
3= Sidechain: hiding + bonding + sidechain
4= Other: density + reference

----------Force zone--------------
Note: A way to unselect zones is to freeze them.
Forced zone: In order to work only on selected zones, type them here.
Walk in the forced zone: all possible zones of related length, only within this selection (if possible)

----------Launch automatically GAB--------------
Allow GAB to be automatically launched after a rebuild+shak+wiggle which gains more than
Launch threshold
(not waiting for further rebuilds, fuzes etc); to save time

Clash rebuild to change clashing importance when rebuilding
Precise bands bands centered on the selected (e.g. worst) zones, thus not fully random
(this will change after a number of loops, see GAB options)
Rebuild best segments, best zones instead of worst zones (used to stabilize on begin, or at the very end when nothing else moves)
Display more information in the log

Joined: 04/11/2013
Thank You

A BIG thanks for making this script understandable. At last I have some understanding of how some of these things work.
Your willingness to share information publicly is commendable. I trust that others will follow your example.
Thanks again.
Porky :)

Joined: 09/24/2012
Groups: Go Science
thks Porky

It encourages me to continue with 3th pannel ...

Joined: 09/24/2012
Groups: Go Science
and the GAB options

When pushing on "GAB", you'll see something like this (some options might change depending on the type of puzzle, e.g. ligands see bellow):

The Ebola GAB strategy:

GAB is started if rebuild gave a sufficient score (see first panel), either immediately after a very good rebuild, either on the best solutions found after a number of rebuild on a focused zone. For each GAB generation, there are a number herds and for each herd, a number of critters to be tested (see bellow). GAB was invented before I came to Fold, by rav3pnl I think. We simple copy-past it to Ebola, like we copy-pasted tvdl's eDRW (+simplifying and harmonizing them).

In Ebola, bands are added from and to accepted segments, from worst segments of the zone in consideration (see first panel) to other random segments, including Use Ligand or not or randomly: "Use Ligand options : 0-Never, 1-Random, 2-Allways when relevant.

Personally, I find great this ability to focus Precise bands on selected area ! It's also useful for Hotspot puzzles.

I don't know exactly the complexity of GAB, but I understand it tries bands one after the other within a group of bands called herds. For each herd, it calculates the score change by each band, keeps only the bands scoring Loss max/crit and go to next critter if under Give up threshold.

The default options (embedded in the recipe - you can edit it) are:
Number of best critters to breed: 3
Number of best critters to keep: 5
Number of best critters to produce bastards:2
minimum bands: 2
maximum bands: 5
Critters with score lower than -50 don’t go to the next generation
Minimal score to do some kids: -20
Bastards always breed
Number of random critters added for each GAB/herd(?) generation: 4

Very important: do you want to Keep conditions (filters etc) before the recipe decides to keep or reject critters?

Fuse threshold determines when to try a Fuze (on, each herd/critter) before to calculate the total GAB generation gain. Remember that here, the Fuze is global (on all protein), thus very slow !!! don't let it to start fuze too often on big proteins ! And use Low Wiggle Power because DRW and GAB need flexibility of the protein.

Don't ask me what all of this represents exactly. I don't know. But it works !

What YOU have to know for Ebola options: full random bands ? or Precise bands centered on the "worst" zones in action? and/or to ligands? Many or few bands (critters) to try? High or low perturbation (critter-accepted loss, GAB accepted loss)?

GAB max loss determines the total loss of GAB that will be accepted before to try a new GAB generation. This is what I change often (big loss takes much more time but it should be advisable on begin or mid game).

This can be endless. that's why we decide to stop GAB after a Max generations without any gain (it's time to try another "worst" zone !). However, if gaining above a certain level, GAB will continue until no gain. When 5-6 "gained another" occur, it's great to see it in the log.
Normal GAB recipe allows an endless number of generations without gain. You may put it high in order to "force" a focused GAB on the latest bad zone on the very end ... if you have days and weeks patience.

If you selected the Precise bands default in the "advanced" panel, it's quite interesting to change strategy after some time and go to a "normal" GAB with Start full random after the given number of generation without gain. And to force Stop if no result after these 2 strategies in order to be able to try other worst zones. Remember that the Start full random works on the all protein like a normal GAB, so be patient and use preferably many Precise bands, stopping immediately after one or 2 additional Start full random.

--------------GAB Options-------------

GAB max loss: Maximum loss per generation;
Max generations: Maximum GAB generations

Max generations without gain for :
Start full random: Abandoning precise bands, go for normal GAB, full random bands
Stop: Stop GAB immediately if no gain after this total number of
bad successive generations (precise and random)

Crits 1r gen.: Number of critters in first generation
Loss max/crit
Give up threshold
Keep conditions: Next critter if a condition is lost
Fuse threshold: Launch fuse after this generation if next to the highscore
(sorry it should be under GAB I think, not here in critters)

"Use Ligand options : 0-Never, 1-Random, 2-Allways
Use Ligand: on ligand puzzles only

Joined: 09/24/2012
Groups: Go Science
Don't worry

I'll try to update Ebola with the "best known today" options whenever possible.
Simply push on "OK" and it should work for overnight run. This is not a recipe to use on hand fold ! It takes many hours to gain!
We almost never use it with these default on begin game: first try simple edrw (or "skip GAB"), then your preferred normal GAB (or all options in order to have a lot of GAB in the GAB panel), then Ebola with this default options.

I'm working on a version that would give you the best default options out of context. But the first priority is stability and best yield in the New Chapter environment.

Joined: 09/24/2012
Groups: Go Science
Recipe outlines

For the worst scoring zones of several segments:
-rebuild and shake several times
-fuze selected for best rebuild_shake results only
-if it's a good solution, add bands from and to this zone, wiggle selected, fuze if gain, wiggle all
-go to next worst zone

Joined: 09/24/2012
Groups: Go Science

Changed shake selected by QuickShake

Joined: 02/08/2012
Groups: None
longest rebuild first

As for as I know, you cant do longest rebuild first. if that function isnt there, can you please add it? thanks

Joined: 09/24/2012
Groups: Go Science

Thanks drumpeter. Added Longest rebuild First in the advanced panel.

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

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