Electron Density (ED) Puzzles

Case number:845813-2001368
Topic:Game: Tools
Opened by:jeff101
Status:Open
Type:Suggestion
Opened on:Wednesday, October 28, 2015 - 13:43
Last modified:Thursday, March 31, 2016 - 07:55

Below are some questions about Electron Density (ED) Puzzles, based on Puzzle 1152:

(1) If you press Tab on a segment to show its Segment Information, does the ED only contribute via the "Density" subscore? Does the ED ever contribute via "Other", "Reference", or any other subscore?

(2) On Puzzle 1152, I have seen the "Density" subscore be nonzero, sometimes positive and sometimes negative, even when the protein is far away from the ED cloud. Is the exact same ED cloud tiled repeatedly through 3D space but only shown for one small region of 3D space? What do negative "Density" subscores mean?

(3) When you press the "Center Protein on Density" button, the overall score you get depends on the "Threshold" setting. Does the overall score depend on any of the other settings in the Electron Density Window (see the image on the left below)?

(4) I would like to play around with the settings in the Electron Density Window while a Recipe is running, but I worry that I will press the wrong thing, change how my protein is scored, and so mess up what the Recipe is doing. Is "Center Protein on Density" the only thing in the Electron Density Window that will directly alter the overall protein score? If I vary things like the "Threshold" but never press "Center Protein on Density", am I only affecting how the ED cloud is displayed? Can selecting "None" as opposed to "Solid" or "Wireframe" affect the overall protein score?

(5) Is it possible to list values for "Red", "Green", "Blue", "Alpha", and especially "Threshold" in the Electron Density Window like you already list for the "Clashing Importance" in the Behavior Window? This way, we could easily tell each other what settings we find work best. Right now, we have to say things like "Threshold to far left", "Threshold at the center", or "Threshold over the m in mouse". It would be easier if we could say something like "Threshold=0.05" instead.

(6) http://foldit.wikia.com/wiki/Electron_Density_Puzzles is helpful. Are there other places you would recommend for help with ED Puzzles? I think many players have trouble seeing things within the ED cloud. It would be nice if there was one place that described every item in the Electron Density Window, explaining things like "Alpha" and "Backface Culling", for example.

Thanks!

(Wed, 10/28/2015 - 13:43  |  26 comments)


Susume's picture
User offline. Last seen 2 hours 4 min ago. Offline
Joined: 10/02/2011

Personally, I recommend never using Center Protein on Density. It is essentially a random move of the protein in space. I have never seen it make the density match better. It does not look at the current position and try to improve on it. I took a protein well-matched to the density (overall density score 3735) and clicked Center Protein on Density, and it moved the protein about 90 degrees and reduced its overall density score to 90 points. The one thing it is useful for is if you have not made any progress matching the density and are having trouble finding the cloud.

It is safe to use the None/Solid/Wireframe etc., Backface Culling, Red/Green/Blue, Alpha, Threshold, and shift-Q controls during a script - they will not alter the score. I tested all of them during wiggle and none of them changed the outcome.

jeff101's picture
User offline. Last seen 23 hours 25 min ago. Offline
Joined: 04/20/2012
Groups: Go Science

Since the results of "Center Protein on Density"
seem to depend on the setting of "Threshold",
perhaps the following would give better results:

(1) Set "Threshold" to the far left,
then press "Center Protein on Density"
until the score stops changing.

(2) Set "Threshold" to the center,
then press "Center Protein on Density"
until the score stops changing.

(3) Do the same thing for
several other "Threshold" settings.

(4) At the end, use the Undo Graph or
Restore Recent, Credit, or Very Best
to get the best possible score.

I'd imagine there is a sweet spot for the
"Threshold" that gives the best results.

If we had LUA commands to read and set the Threshold
(for example, density.GetThreshold and density.SetThreshold) and
press the "Center Protein on Density" button (density.CenterProtein),
we could make Recipes to do the above procedure automatically.

LociOiling's picture
User offline. Last seen 31 min 37 sec ago. Offline
Joined: 12/27/2012
Groups: Beta Folders

I find the density cloud has a tendency to disappear suddenly when I try to rotate it.

Trying to view the density, I move the protein far away. Then I use "Focus on Density" from the Electron Density dialog. When I try to rotate the density in any direction, it quickly fades and then disappears. Often it seems to just disappear.

It would be nice if there was a mode where you could move the density as you would the protein. It would seem like "Focus on Density" could select this mode, and then a new "Focus on Protein" would switch back to normal mode. As it is, it seems like there's some mystery center point relative to the protein that causes the disappearing density issue.

Just in general, it seems like a "drone mode" would be a nice option. In other words, treat the protein and the density as floating motionless in space, and have the "camera" fly around them. The foldit model seems to hold the camera still and move the protein around. There's always a center point, which becomes an issue when things are spread out. An option to show the center point would also help in this regard.

My next step is to make a compact ball of the protein and move it just a little to the side of the density.

LociOiling's picture
User offline. Last seen 31 min 37 sec ago. Offline
Joined: 12/27/2012
Groups: Beta Folders

Even with a reasonably compact protein off to one side of the density, the density tends to disappear when you rotate.

The best solution I've found so far is to switch to the "Line" view with the EnzDes or CPK coloring, still with an extended chain. Then "Center Protein on Density" from the Electron Density panel and maybe "Focus on Density" if needed.

These settings minimize the visual distraction of looking at the protein, and let you rotate the density without having it disappear.

Susume's picture
User offline. Last seen 2 hours 4 min ago. Offline
Joined: 10/02/2011

To return from focus on density to focus on protein, type Q.

I always place the extended chain in the cloud to examine the cloud at the beginning. This lets me use shift-Q to zoom in on portions of the cloud and to see the interior. Place the extended chain in or near the portion you want to examine, hover mouse over backbone, and hit shift-Q. To examine another area, shift the extended chain to or near that area and do it again. I never attempt to see into the cloud without having some protein in it to anchor the focus to.

Joined: 09/24/2012
Groups: Go Science

Excellent tip !! Thanks ! I'm now a little bit less blind to the interior of these clouds.

LociOiling's picture
User offline. Last seen 31 min 37 sec ago. Offline
Joined: 12/27/2012
Groups: Beta Folders

Hotkeys and shortcuts for electron density in case I forget again.

Thanks Susume for mentioning shift-q. When you hover on a segment, shift-q limits the density display to things in the same plane that segment. Hitting shift-q again (or just plain q) restores the full view of the density.

Shift-q does not seem to be present in the foldit help (under Menu->Help).

The near and far visibility can also be helpful.

The shortcut ctrl-shift-click-drag changes the far distance cutoff. So if you hold ctrl+shift, then click somewhere and drag the pointer downward, you'll see the more distant parts of the protein fade and disappear. If you ctrl+shift then click and drag the pointer upward, the distant parts reappear.

The shortcut alt-shift-click-drag changes the fog or haze. If you hold alt+shift, then click and drag downward, the more distant parts of the protein grow hazy. Dragging the pointer upward cuts through the haze. Unlike the far distance cutoff, the distant parts don't disappear, they just get dimmer.

The fog shortcut doesn't seem to be present in the foldit help (under Menu->Help).

The short ctrl-alt-click-drag changes the near distance cutoff. If you hold ctrl+alt, then click and drag the pointer *upward*, you'll see the near parts of the protein fade and disappear. Dragging the pointer downward brings the near parts back into view. Note that the direction of dragging seems to be reversed relative to the two far view shortcuts, but perhaps there is some higher-order symmetry at work here.

The "q" shortcut seems reset the visibility and fog options, in addition to clearing shift-q and potentially moving the protein back to front and center.

And of course the shift-e shortcut toggles the Electron Density panel.

Joined: 01/17/2012
Groups: Contenders

Loci the shortcut tips were VERY helpful. Thanks

alwen's picture
User offline. Last seen 5 weeks 4 hours ago. Offline
Joined: 10/03/2011

Shift-q and the fog shortcut are found in Help in the upper left, in the little square box that says "IRC Commands" at the top.

If you use the scroll bar to view the rest of that box, you will find

Camera Controls
Mouse on the background
Adjust fog
and
Mouse on a segment

I agree that they are hard to find in Help!

LociOiling's picture
User offline. Last seen 31 min 37 sec ago. Offline
Joined: 12/27/2012
Groups: Beta Folders

Added a slightly improved version of the above to the wikia hotkeys and shortcuts page: http://foldit.wikia.com/wiki/Shortcut_Keys#Visibility_Shortcuts

Also updated the electron density puzzles page a bit with a reference to 1152 and some other fun facts: http://foldit.wikia.com/wiki/Electron_Density_Puzzles

jeff101's picture
User offline. Last seen 23 hours 25 min ago. Offline
Joined: 04/20/2012
Groups: Go Science

http://fold.it/portal/node/2001344#comment-31344 has images of the 20 amino acids and how they might look in the electron density cloud.

Susume's picture
User offline. Last seen 2 hours 4 min ago. Offline
Joined: 10/02/2011

A numeric scale on the trim dialog would be very helpful, as would making the trim slider stay put from one viewing to the next. Currently the slider is reset every time you hit the trim button, so you can't increase or decrease relative to the current setting, because you can't see the current setting. Trim slider could possibly be added to the regular ED menu rather than being a button and separate dialog.

frood66's picture
User offline. Last seen 3 hours 10 min ago. Offline
Joined: 09/20/2011
Groups: Marvin's bunch

as far as I understand it - this function works on rmsd - so good luck with that.

jeff101's picture
User offline. Last seen 23 hours 25 min ago. Offline
Joined: 04/20/2012
Groups: Go Science

Susume has made a new ED puzzle video. For more info see:
https://fold.it/portal/node/2001427

Thanks again, Susume!

jeff101's picture
User offline. Last seen 23 hours 25 min ago. Offline
Joined: 04/20/2012
Groups: Go Science

I just watched Susume's latest ED puzzle video and had this idea:

It would be neat if instead of just putting a red dot, you could also fill in with red the region of the cloud where you put the red dot, kind of like in Paint programs. The region filled would depend on the threshold setting, so it would be possible to fill just small regions of the cloud if you wanted to. Regions not connected to where you put the red dot would not be filled.

If you don't like how your filled red region looks, you could delete it all at once or just erase part of it.

Perhaps you could place borders within the cloud and then fill just the region between the borders.

jeff101's picture
User offline. Last seen 23 hours 25 min ago. Offline
Joined: 04/20/2012
Groups: Go Science

Another idea is to give players an option to turn off the density subscore so they can fold the protein without any influence from the ED cloud. Once the protein is fairly well folded in this way, the player could turn the density subscore back on and then try to put the folded protein into the ED cloud as best they can.

Joined: 09/24/2012
Groups: Go Science

Excellent idea ! It's a shame to try to put a externally done structure (or imperfect) inside of the cloud, and the solution being distorted by the density bonus.

Another way to do it: using edrw recipe with no density in the slots neither worst list.

New scripts with options "score without density" when calculating the score is also an option (but too much work adding this to all recipes !).

jeff101's picture
User offline. Last seen 23 hours 25 min ago. Offline
Joined: 04/20/2012
Groups: Go Science

Another way to implement it would be as below:

Say the ED cloud when made visible only occupies a cube with sides of length L.
Have the density subscore contribute to the total score of the protein
only within a cube with sides of length 5L centered on where the ED cloud appears.
Outside this cube with sides of length 5L, have the density subscore always be zero.
This way, if you don't want the density subscore to contribute to the total protein score,
just move the protein far from where the ED cloud appears.

jeff101's picture
User offline. Last seen 23 hours 25 min ago. Offline
Joined: 04/20/2012
Groups: Go Science

There could even be a button to move the protein outside the cube with sides of length 5L.

Joined: 09/24/2012
Groups: Go Science

I saw the latest great video from Susume. But you need quite chance to find the sheets and other macro-structures with this method.

What would be great is the following tool:

-Like for medical scanner navigation: a slice navigation tool. You have a 2D slice (of even a tick slice or ships) that crosses the density and you can move it forward or backward. You don't see anything else than this slice. You don't need any segment inside of the cloud if you ant to explore the cloud (nice if you build your SS outside the cloud and you want to explore the cloud on mid game).

It would be more quick than Shift-Q then hide backward and front. Then try again with another shift-Q (where you need always some segment to focus on).

-Option completely hiding the puzzle (not the cloud), like we can hide the cloud not the puzzle.

Joined: 09/24/2012
Groups: Go Science

and this would also be great:
http://fold.it/portal/node/2001413

and/or a tool that could help "putting the sheet into its cloud":
Shif-Q on a segment (of the sheet for example). Then you can anker the slice on this segment. When you move the segment, the focus follows the segment - or better the number of selected segments (as if you move and shift-Q in parallel).

Joined: 03/30/2013
Groups: Go Science

I wish it were easier to navigate the Z direction.

Susume's picture
User offline. Last seen 2 hours 4 min ago. Offline
Joined: 10/02/2011

This inspired me to play around with the controls. Place a small piece of protein in the cloud. Mouse over it and hit Shift-Q to give it the focus. Bring up the Move tool on that piece. Use Shift-drag on the purple tool to move the piece in the Z dimension - the camera focus will follow it into and out of the screen, letting you move forward and back through the cloud. Not great, but maybe it's better than having to roll 90 degrees and back to move through that dimension.

frood66's picture
User offline. Last seen 3 hours 10 min ago. Offline
Joined: 09/20/2011
Groups: Marvin's bunch

Indeed it does...I have used this method for some time (think I referred to it last night when discussing yr excellent vid Subsume)

jeff101's picture
User offline. Last seen 23 hours 25 min ago. Offline
Joined: 04/20/2012
Groups: Go Science

In the most recent Developer Chat
(http://fold.it/portal/node/2002210)
from 15:10-15:21, bkoep and jflat06 discussed
letting players "trim unwanted residues" in
electron density puzzles so that their electron
densities would match the experimental electron
density better.

If the trimmed residues are at the end of the
protein chain, this seems fine to me, but if
the trimmed residues are in the middle of the
protein chain, things get more complicated. If
you remove the trimmed sidechains, maybe you
lose some hydrogen bonds or create more voids
in the protein. If you remove the trimmed backbones
too, you might need to apply some constraints so
that the remaining parts of the protein chain stay
a reasonable distance apart.

Perhaps a better way is to make like an
invisibility cloak around the residues you want to
trim. The trimmed residues are not really removed
from the protein chain. Instead, the trimmed residues
behave like they normally do and give normal
Rosettta/Foldit scores, but the invisibility cloak
sets all of their calculated electron densities to zero.

jeff101's picture
User offline. Last seen 23 hours 25 min ago. Offline
Joined: 04/20/2012
Groups: Go Science

In the "Electron Density" menu above, would it be possible to have several options as below:

One option (the default) shows both the protein and the experimental electron density the normal way, based on settings in the "View" menu for the protein and based on settings in the "Electron Density" menu for the electron density (ED).

Another option makes the experimental electron density disappear and calculates what ED the present protein conformation would give. It then renders the calculated ED the way it would normally render the experimental ED.

Perhaps a 3rd option would make the present protein conformation disappear and would render both the experimental and calculated ED's the normal way, based on the settings in the "Electron Density" menu. If the experimental and calculated ED's could be shown with different shades or colors, I think that would help.

With options like these, players could force the protein into all beta-sheet and then see what the ED would look like for all beta-sheet. Similarly, players could try all alpha-helix and then see what the ED for all alpha-helix looks like. Players could force disulfide or hydrogen bonds to form and then calculate what the ED would look like. I think this would open new ways for players to do ED puzzles.

Ideally, the closer the present protein conformation is to the desired one, the more similar the calculated ED will be to the experimental ED.

Thanks for reading!

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