Electron density

Case number:699969-989897
Opened by:marie_s
Opened on:Saturday, June 25, 2011 - 09:33
Last modified:Saturday, September 22, 2012 - 09:04

New electron density in beta version - see in faq how to test beta version.

Alignment is time consuming.

Great graphics options but, for my taste,too close to QTTN to be fun on this puzzle.
I would like to try it on freestyle puzzle.

Crash once but not reproducible.

(Sat, 06/25/2011 - 09:33  |  6 comments)

jflat06's picture
User offline. Last seen 2 days 3 hours ago. Offline
Joined: 09/29/2010
Groups: Window Group

Thanks for the reply.

You're right that this puzzle is pretty close to a QTTN. In this particular puzzle, the electron density isn't from actual data, and has been derived from the native structure itself. Puzzles which include real experimental data will have less precise guides, and you'll have to really think about which pieces of the protein go where.

The alignment taking a while is something we know about, and we'll try to speed things up a bit for you all.

Joined: 12/16/2010

banding to 438 ED guide is possible

It is possible to band to the Electron Density guide of puzzle 438.

Thanks go to Marie, global chat involving BootsMcGraw and others plus discussions and key support in AD group chat for inspiration, advice and encouragement.

See link in my earlier reply (below) to image suggesting View settings for banding ED.

Not counting false starts, it took maybe five hours to create this complete banding.

So here's an image of the complete banding of a fresh reset. I banded to the original alignment in one long go (of the foldit client -- there were real-life breaks!), but the guide may have moved a bit since.


That banding was successfully saved and shared to group. Save/Share delivers band geometry intact. The key fact is that this was done at all. I expect the art of banding to ED clouds will advance rapidly as the more challenging ED puzzles are presented.

This is what the model looks like after I apply wiggle at CI=0.12 until the score reduces to about 6858.


On my first try using the banding I really slammed the model into strongly negative scoring. After wiggling back up to around 10172 I abandoned the bands and guide and applied a couple of the usual scripts. This got me to within about 17 points of my previous best. For the following image I used the Move Tool to approximately align the result with the guide. There seems to have been considerable drift, but the result still looks to be in the general neighbourhood of the guide (at least if you squint a bit).


Banding to this Electron Density cloud was quite an adventure. Here are some rough and ready tips from my first pass:

1. use initial alignment and don't touch the Move Tool until banding complete

2. Band from one structure at a time, doing the easy cases first and then using these as aids to interpolation for the hard ones. I ran Recipe #16511 "Report on Structures v0.03" and copied the output to a text file as an aid for organizing this.

3. When banding from a new structure ...
a) hover/Tab over the first segment to bring up "Segment Info" as a pointer
b) Disable Bands(B) to fade out bands already placed
c) make very approximate placement of bands into the cloud for each segment in structure
d) perform View --> View Protein: Line to minimize model image
e) bury band tips from structure into the appropriate spots in the cloud

From my log here's an extract describing which bits of the cloud I found difficult:

Banding close / relatively obvious structures:

Loop Alfa(1,2)
Sheet Bravo(3,7)
Loop Charlie(8,11)
Sheet Delta(12,20)

... the dark zone

Sheet Papa(69,76)
Helix Quebec(77,80) -- **dramatic twist at 77 **
Loop Romeo(81,86) -- ** interpolated guesswork **
Sheet Sierra(87,90) -- **choppy**
Loop Tango(91,94)
Sheet Uniform(95,101)
Loop Victor(102,105)
Helix Whiskey(106,123)
Loop Xray(124,133)
Sheet Yankee(134,138)
Loop Zulu(139,147) -- **tricky**
Sheet 27(148,154)
Loop 28(155,157)

Xtreme banding (segs 21 - 68)

Loop Echo(21,22)
Helix Foxtrot(23,25)
Loop Golf(26,29)
Helix Hotel(30,38) -- **translation ahead, a tough move**
Loop India(39,42) -- **random cloud-filling interpolation**
Helix Juliet(43,48) -- **location and helicity both iffy**
Loop Kilo(49,55) -- **large displacement and guide ambiguous here**
Sheet Lima(56,60) -- **key move, I put it ahead about 1 seg**
Loop Mike(61,61)
Sheet November(62,63)
Loop Oscar(64,68)

Joined: 05/18/2008
Groups: None

Alignement toolsdesn't work well after big moves with the purple star.
See pictures in the wiki http://foldit.wikia.com/wiki/Puzzle_438

Joined: 12/16/2010

Hi Marie,

Tried to add the following picture to your page (suggested View ED settings) and succeeded in erasing existing pictures. :-( Think I have undo'd my mistake now. Will attempt add later.


jflat06's picture
User offline. Last seen 2 days 3 hours ago. Offline
Joined: 09/29/2010
Groups: Window Group

The alignment tool is much less precise than the one you have for QTTN puzzles. It wont be able to sort out how to align things if you make big changes like that. If you find yourself unable to align, you may have to do a bit of manual alignment with the purple star before using the alignment.

Joined: 05/18/2008
Groups: None
Status: Open » Closed

new feedbacks on this


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