[Chimera-users] FAD binding site on Cryptochrome

Elaine Meng meng at cgl.ucsf.edu
Thu Oct 10 15:46:42 PDT 2019

Hi Joseph,
Wow, I have to say it is a project in itself just to understand the different output files from this server!  I still don’t understand most of it, but I’ll try to get you started.

(1) The most straightforward part is that you can certainly open a PDB file and display and label any positions that you want. If you know the residue numbers of the Trp residues, you can just specify them in that way, e.g.

display :52,148,62,250
rlabel :52,148,62,250

…to show atoms and labels for residues with those numbers  Or, you could select the residues in the sequence (menu: Favorites… Sequence) and then use the Actions menu to display and show labels. These labels move/rotate together with the structure.  However, for publication images you might instead want 2D labels that don’t move with the structure, but are drawn on top (analogous to like adding labels afterward in Photoshop). There are too many aspects of making nice images to list here but here’s the page of “image tips”:

If you don’t know in general how to hide, show, specify … then perhaps should take a look at one or the other of the “getting started” tutorials:

(2) Showing the probability values on the residues.  I downloaded example output files from the COACH server and see that the probability text file lists the residue numbers and the probability values.  You can reformat this into an “attribute assignment file” (also a fairly simple text file) and after opening the corresponding structure, assign the values and then use the Render by Attribute tool to show the values with continuous colormaps (like is often done to show B-factors) and/or backbone “worm” thickness.  The  example file downloaded from COACH looks like this

r=MET;n=1 :prob=0.026
r=HIS;n=2 :prob=0.026
r=HIS;n=3 :prob=0.026

and the attribute assignment file could be like this

#  Use this file to assign the attribute in Chimera with the 
#  Define Attribute tool or the command defattr.
attribute: probability
match mode: 1-to-1
recipient: residues
	:1	0.026
	:2	0.026
	:3	0.026

… where there are tabs before both columns, format described here along with example files

You would start Define Attribute (menu: Tools… Structure Analysis… Define Attribute) to read in the file and then use the Render by Attribute dialog that pops up to do the coloring and/or “worm” display.

If you really want to label by the values as opposed to showing with color or worms, you could also do that after loading the attribute, using “labelopt info”

(3) labeling the pocket residues.  It looks like the “Bsites.inf” example output from COACH gives a bunch of numbers for each site, and I suspect that at least some of these are the residue numbers, i.e. which protein residues to display/label.  Actually, the web page showing the results also gives a list of “Consensus Binding Residues.”

Example results page
shows for the first-ranked prediction, residues 72,98,99,100,172,173,174,192,193,194,195,217,218,228

In Chimera, I opened the downloaded Rep Templ file for that same row of results and used commands:

display :BIG
display :72,98,99,100,172,173,174,192,193,194,195,217,218,228
rlabel :72,98,99,100,172,173,174,192,193,194,195,217,218,228

… this did indeed display and label the residues that appeared to be near the ligand named BIG.
I wasn’t sure if the protein in the file was the same as the “3D structure model” but they seemed to have the same numbering as each other, and the same residues in the same positions.  You’ll have to doublecheck that whatever downloaded PDB you display really contains the protein that you expect it to, since it is not clear to me from the information on this page.

(4) comparison with cluster.  Looks like the “Bsites.clr” example output from COACH gives, for each cluster, a series of PDB IDs including chain ID, and for each of those, several numbers I don’t understand but the last set looks like it might be residue numbers.  (However, I don’t know if the residue numbers refer to the predicted structure or to the numbering in that other PDB file.)  So in addition to the predicted structure, you could also open some of these other PDB structures and superimpose them.  You would probably want to delete the irrelevant chains from those other structures. 

There are a number of approaches for superposition.  Matchmaker will try to figure out everything for you, but to precisely control and specify which residues should be used in fitting, then you’d need to use the “match” command instead.  Superposition methods are discussed here, with links to the details for each:

The example has 780 structures in one cluster, which would be too many to superimpose.  In theory, you could superimpose them all but unless they are all extremely similar, it would look like a pile of spaghetti.  So you would need to consider what you really want to do…   

I’ve probably already given more nitty gritty than you really want to digest in a single message.  If you get stuck or have questions on something specific, please feel free to write back.
I hope this helps,
Elaine C. Meng, Ph.D.
UCSF Chimera(X) team
Department of Pharmaceutical Chemistry
University of California, San Francisco

> On Oct 10, 2019, at 9:34 AM, Margiotta, Joseph <Joseph.Margiotta at utoledo.edu> wrote:
> Hi->
> I am very new to molecular modeling and have a very basic novice user issue.  We are studying cryptochrome proteins some of which are known to bind the chromophore Flavin adenine dinucleotide (FAD) and for this and other reasons are considered photopigments.  We have looked at a cryptochrome in chicken that binds FAD and is light sensitive (GallusCRY4) and have sent the AA sequence to Zhang’s COACH site for Protein-ligand binding site prediction (https://zhanglab.ccmb.med.umich.edu/COACH/) at the University of Michigan.  The results come up with a C-score confidence prediction of 0.59 with a cluster size of 63 for FAD binding to GallusCRY4.  The results allow you to download the 3D model and the residue-specific binding probability, which is estimated by SVM, the predicted bound ligands and detailed prediction summary, and the templates clustering results. We’d now like to extend this modeling to GallusCRY1 and GallusCRY2 to see get a FAD binding prediction and the complex structure with the most representative ligand in the cluster.
> The COACH site lets you export the model and complex structure in PDB format and I downloaded Chimera to visualize and analyze the PDB results.   I would like to be able to label the structures using the binding probabilities, and template clustering results to include the predicted FAD binding pocket on the three CRY proteins and also label a tetrad of Trp residues that are thought to be responsible for electron transfer.  Can you give me some general guidance on how to start this analysis? 
> -- Joseph F. Margiotta, PhD
> Professor of Neurosciences
> UT College of Medicine & Life Sciences
> Block HS 120A   Mail Stop 1007
> Toledo, OH  43614
> 419-383-4119

More information about the Chimera-users mailing list