Electron Density Puzzles

Case number:845813-2006117
Topic:Game: Tools
Opened by:jeff101
Opened on:Saturday, October 20, 2018 - 05:29
Last modified:Thursday, October 25, 2018 - 00:37

I've been working on the Electron Density Puzzle 1588 tonight.
One goal was to put purple dots in the cloud where I thought
the sidechains were and red dots in the cloud where I thought
the backbone was. It seemed like if I tabbed on a part of the
cloud, it put a purple dot there. If I wanted a red dot, it
seemed like I had to right-click on a purple dot and then
select red for its color. Is it possible to click something to
change the default dot color to red so that the next series of
tabs will make red dots instead of purple ones? Then I could
pick a default color, tab to make a series of dots with that
color, then pick a different default color, then tab to make
a series of dots with this new color. If this isn't possible
now, can you change the interface so it will work in the

I asked in Chat about the above, and Loci requested yellow
as a dot color option as well. Then we could pick among
red, yellow, green, blue, and purple for our dots' colors.

(Sat, 10/20/2018 - 05:29  |  12 comments)

jeff101's picture
User offline. Last seen 6 hours 35 min ago. Offline
Joined: 04/20/2012
Groups: Go Science

Another question I had was if there was a page on the
Foldit website or in its wiki showing examples of how
different sidechains looked in the electron density
cloud in actual Foldit puzzles. The images at
https://fold.it/portal/node/2001386 are useful, but I
think more sample images from within actual Foldit
puzzles would help.

Would someone able to recognize the different
sidechains post images of each one rendered several
different ways (Solid, Wireframe, Isocontours, and
Grid)? If there is variation from puzzle to puzzle
or even within the same puzzle, would someone post
additional images to help players learn to recognize
the different shapes?

Thanks for reading!

frood66's picture
User offline. Last seen 9 hours 10 min ago. Offline
Joined: 09/20/2011
Groups: Marvin's bunch

it depends on the level of definition - but, using wire frame mode (my personal favourite) I don't believe I could improve on the images u mention.

jeff101's picture
User offline. Last seen 6 hours 35 min ago. Offline
Joined: 04/20/2012
Groups: Go Science

How about the proline? I would expect the backbone to bend
at each proline, but the image cited above doesn't really
show the backbone. Can you actually see proline-induced
bends in Foldit's electron density cloud?

frood66's picture
User offline. Last seen 9 hours 10 min ago. Offline
Joined: 09/20/2011
Groups: Marvin's bunch

on 2nd inspection - it would, perhaps be better if the images u refer to showed the backbone also. I also see yr point re proline. But, in truth, prolines are quite easy to recognise - not so much by the backbone bend but more by the side chain 'blob' being large and close/part of the backbone cloud. Ultimately, the most important thing is the definition - the higher the better. This puzzle does not have the highest definition we have seen - but it is good enough. In any event, the viewing settings are paramount. again, this puzzle's default required no changes for me - but it is very much a personal thing.

jeff101's picture
User offline. Last seen 6 hours 35 min ago. Offline
Joined: 04/20/2012
Groups: Go Science

Another idea for puzzles like 1588 with multiple starting
structures would be to let players copy all the dots & notes
they build in the electron density cloud from one starting
structure or saved solution to another.

Joined: 12/27/2012
Groups: Beta Folders

I'd also like a way to keep the density note dots around when doing a reset.

frood66's picture
User offline. Last seen 9 hours 10 min ago. Offline
Joined: 09/20/2011
Groups: Marvin's bunch

yes please - keep those density notes on reset

jeff101's picture
User offline. Last seen 6 hours 35 min ago. Offline
Joined: 04/20/2012
Groups: Go Science
Another thing I'm curious about is what are good 
Density subscores for each type of amino acid? 
For example, if you tab on Methionine 23 in 
Puzzle 1588 and it gives a Density subscore 
of 18.6, is that a good result for Methionine? 
Is  7.8 a good result for Lysine         2?
Is 15.8 a good result for Threonine      4?
Is  5.6 a good result for Glycine       13?
Is 10.4 a good result for Alanine       22?
Is 18.6 a good result for Methionine    23?
Is 18.5 a good result for Tyrosine      24?
Is 13.3 a good result for Asparagine    25?
Is 20.2 a good result for Glutamate     27?
Is  5.8 a good result for Glutamine     48?
Is 13.8 a good result for Arginine      51?
Is 11.1 a good result for Valine        58?
Is 19.1 a good result for Aspartate     67?
Is 10.2 a good result for Isoleucine    77?
Is 16.4 a good result for Proline      170?
Is 15.7 a good result for Leucine      177?
Is 13.1 a good result for Serine       188?
Is  7.3 a good result for Tryptophan   197?
Is 15.3 a good result for Phenyalanine 212? 
What are good results for Cysteine & Histidine?
What does a negative Density subscore mean? 
Is a negative Density subscore undesirable?
Are there known maximum possible Density
subscores for each amino acid that would
be the same for every puzzle?

Is there a recipe out there that lists each 
segment's amino acid and its Density subscore? 
I'd be curious what ranges are typical for each 
amino acid.
Joined: 12/27/2012
Groups: Beta Folders

print protein 2.8 includes some density analysis, and at least lets you compare the amino acids in a particular pose. There's cut-and-paste report that shows the mean density score by amino acid, along with the best and worst densities identified by segment number.

With persistence, you could accumulate the results across puzzles, but I haven't tried that.

Joined: 12/27/2012
Groups: Beta Folders

As also mentioned in chat, I found that electron density info for many proteins is out there on the internet. There's a "PDBe", which is separate from the PDB hosted at rcsb.org.

There's a LiteMol viewer which can display density along with a protein. It's available at http://webchemdev.ncbr.muni.cz/Litemol/Viewer/

The user interface is a little confusing, but there's tree structure on the left. The options for the selected node of the tree are shown on the right.

You start with the "root entity" node, enter the PDB id of the protein you want under "molecule" on the right, and click "add". (PDB 1CBS is there as a suggestion.) That gets you the typical protein viewer display.

To see the density, re-select "root entity", and again enter the PDB id, this time under "density data from PDBe". When you click "add", the density gets downloaded and processed.

By default, density is displayed around a selection, so you have to click on the protein to see anything. When you see density, there are several different types displayed. Some density is represented in a solid view, some is wireframe.

You select the nodes from the tree structure and adjust how different types of density info are displayed. You can also hide density using the "eyeball" on a node.

I haven't deciphered all the density types, but I believe the main idea is comparing the density data to what you'd expect from the protein model. The greek letter σ (sigma) appears in various spots, which I suspect refers to standard deviation units. I'm guessing it's similar to the threshold setting in Foldit, a higher sigma would be like a lower ED threshold.

There's also "Fo-Fc" and "2Fo-Fc" density, again assuming those involve comparison, but I'm not clear on the differences, but "Fo-Fc" gets wireframe by default, and "2Fo-Fc" is solid.

Bottom line, LiteMol is oriented toward assessing how well a protein model and its density match. That's a little different than what we do in Foldit, trying to fit the protein to the density. Still, it seems like the option to display density within a certain distance of a selection would be useful in Foldit. It's a little similar to shift-Q, but the protein and density seem less likely to go out of focus as you rotate.

There are lots of other settings and options in LiteMol, and other viewers may have similar functions. Further investigation pending....

See also:
*Electron density map (EDM) - http://proteopedia.org/wiki/index.php/Electron_density_map
*Interactive EDM demo - http://www.bioinformatics.org/molvis/edm/

rmoretti's picture
User offline. Last seen 12 hours 16 min ago. Offline
Joined: 01/15/2010
Groups: Foldit Staff

You're correct about sigma. It does have to do with confidence levels (standard deviation). By varying the sigma threshold, you can look at just the parts of the density we're most sure of versus the bits of the density which are a bit more fuzzy.

The "Fo-Fc" is a "difference map". That is, it looks at the difference between the observed density (represented by "Fo") and the computed density ("Fc"). This shows contours of "positive density" where there's experimental data not represented by the model and "negative density" where the model predicts there should be density, but where no density is observed in the experiment. This can be used to home in on regions of the model which aren't representing the data well. (The method by which one calculates an Fo-Fc map is somewhat complicated and time consuming, which is why Foldit doesn't provide it as an option.)

The "2Fo-Fc" is an attempt to combine the Fo-Fc map with a "standard" electron density cloud. Whereas an Fo-Fc map will show no density where the model matches the experimental data, a 2Fo-Fc map will show a "normal" density cloud. But adding in the Fo-Fc component will help highlight/adjust the portions of the density where the model isn't matching the experimental data.

I've haven't read it in full, but https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5051661/ seems to be an open-access paper which attempts to do an introductory-level description of electron density maps.

And for the really enthusiastic, Gale Rhodes's "Crystallography Made Crystal Clear" is an excellent introductory textbook on X-ray crystallography.

Though I should clarify that many of the electron density clouds we're providing with Foldit are produced via Electron Microscopy instead of X-ray Diffraction. At a superficial level these densities are similar, but details are quite different in how they're produced. (For example, technically there's no such thing as an Fo-Fc or 2Fo-Fc map for EM density. There's similar things you can calculate, but Fo-Fc and 2Fo-Fc are strictly speaking just for diffraction experiments.)

jeff101's picture
User offline. Last seen 6 hours 35 min ago. Offline
Joined: 04/20/2012
Groups: Go Science

In Foldit, when you tab on a particular residue and the
"Segment Information" appears, what does a negative Density
subscore mean? I assume a positive Density subscore is
better. Do different amino acids have different maximum
possible Density subscores that would be the same in all
Foldit puzzles? Are these maximum values known? Can you
post a chart of them somewhere?

Also, do you know of any programs that will calculate model
electron densities (Fc above?) for a given pdb file? Should
the model electron density be the same whether one measures
it with Electron Microscopy or X-ray Diffraction?


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