[Chimera-users] electrostatic potential based on Swiss-Model

Elaine Meng meng at cgl.ucsf.edu
Fri Jun 6 13:47:33 PDT 2014

Hello Jingga, 
Your phi map from DelPhi (or dx map from APBS) contains a 3D grid of points, with an electrostatic potential (ESP) value at each point.  The surface coloring just takes the values from that file, interpolates to find values at or near the surface, and then does a value->color assignment.  So, those numbers just correspond to the ESP values read from the grid file.

When you change the values in the coloring dialog, you are only changing the coloring: the map stays the same!  For comparing different structures I would use the same value->color pairings for each image.  Usually the default values for ESP coloring work well (defaults in Electrostatic Surface Coloring or Coulombic Surface Coloring dialogs, both +/-10, I believe).  If you change those, at least use the same values for the different structures. Using "full range" is not recommended, it would give different colorings for different structures because they wouldn't have the same min and max ESP values as each other.

However: it's not possible to tell how well you reproduced somebody else's ESP result by just looking at their paper figure, unless you know what coloring settings they used (including the value->color pairings), and you try those same settings.

Another caution is that I would not rely too much on the modeling score.  I don't know anything about Swissmodel score, but in general, model-evaluation scores don't capture everything about a structure, and a model might get a good score by one method but a bad score by some other method.
Elaine C. Meng, Ph.D.                       
UCSF Computer Graphics Lab (Chimera team) and Babbitt Lab
Department of Pharmaceutical Chemistry
University of California, San Francisco

On Jun 6, 2014, at 9:21 AM, Jingga Inlora wrote:

> Hello Elaine,
> The Swissmodel gave me a relatively good score on the predicted protein (0.73), so I think the problem is most likely #2. I am using default.crg and default.siz files to calculate the electrostatic potential. I have also tried using the coulombing surface coloring as well as APBS server but the results look worse. I've tried it with my other solved protein structure too and wasn't able to repeat some of their reported electrostatic potential maps. I have one general question about assigning colors on the electrostatic map: what does the value of -10 (red), 0 (neutral), 10 (blue) really represent? I am asking this because it seems that if I assign different values to the color, I get a very different electrostatic potential map. Also, if I use my previously calculated .phi files to draw the map and set the values to "full range surface values", the numbers I get vary from protein to protein and I am not sure how to compare them. Do you think I should show them as is, or define a value for all proteins (eg -30 to +30) and compare them?
> Thanks a lot.
> Jingga
> On Thu, Jun 5, 2014 at 7:25 PM, Elaine Meng <meng at cgl.ucsf.edu> wrote:
> Hello Jingga Inlora,
> It is impossible to tell  what (if anything) might be wrong with your structure from Swissmodel or the resulting phi map from DelPhi.  You can't judge just from which program built the structure (in this case Swissmodel), because it also depends on what you used as input to that program and whether similar protein structures are available to use as modeling templates.  I guess you could ask the Swissmodel developers but they would probably give a similar answer.
> This is not really a Chimera question, but I will try to answer generally. If I had the problem of suspicious-looking electrostatic potential, here is what I would check:
> (1) does the atomic structure itself look bad or strange?  Maybe Swissprot wasn't able to make a good model in this specific case, for example if there were no good template structures available.  There are programs that can evaluate the quality of protein structures, for example: Molprobity, PROCHECK, and probably many others.
> (2) was the charge, radius parameter assignment for DelPhi successful?  Sometimes the atom names don't match what is in the DelPhi charge and radius files, resulting in a DelPhi calculation that thinks the protein has a ridiculous large negative charge.  Or maybe you didn't even add hydrogens.  Some of the charge/radius files don't require hydrogens, but many do. In my experience, depending on what DelPhi charge and radius parameter files you are using, it can be very difficult to fix up all your atom and residue names to correctly match.
> if you think the atomic structure looks reasonable (#1 is OK) and the problem is more in #2 above, you can try something easier like the "Coulombic Surface Coloring" tool in Chimera and see whether the result looks more reasonable:
> <http://www.rbvi.ucsf.edu/chimera/docs/ContributedSoftware/coulombic/coulombic.html>
> This tutorial includes Coulombic calculations, as well as the slightly more complicated approach with PDB2PQR and APBS (Poisson-Boltzmann calculations, more similar to DelPhi):
> <http://www.rbvi.ucsf.edu/chimera/docs/UsersGuide/tutorials/surfprop.html>
> Best,
> Elaine
> ----------
> Elaine C. Meng, Ph.D.
> UCSF Computer Graphics Lab (Chimera team) and Babbitt Lab
> Department of Pharmaceutical Chemistry
> University of California, San Francisco
> On Jun 5, 2014, at 1:07 PM, Jingga Inlora <arolni at umich.edu> wrote:
> > Hello,
> > I am a graduate student at the University of Michigan and I am currently trying to obtain images of my proteins with electrostatic potentials for my manuscript. For one of my proteins, there is no solved protein structure (by X-ray crystallography or NMR) and I had to use SWISS-MODEL to predict the structure. I did electrostatic potential calculation using DelPhi but when I visualized it using Chimera, the electrostatic distribution is not what I expected. I did notice the electrostatic potential calculations for other previously solved protein structures took much longer and resulted in bigger files, whereas the SWISS-MODEL-predicted protein contains much smaller .phi file size.
> > I was wondering if you have any idea how accurate it is to calculate the electrostatic potential based on a SWISS-MODEL predicted protein structure, and if there are ways to improve the reliability of the calculated electrostatic potentials for that protein?
> > Thank you very much for your help!
> > Sincerely,
> > Jingga Inlora
> _______________________________________________
> Chimera-users mailing list
> Chimera-users at cgl.ucsf.edu
> http://plato.cgl.ucsf.edu/mailman/listinfo/chimera-users

More information about the Chimera-users mailing list