[chimerax-users] The ChimeraX-Modeller interface

[Elaine Meng]

Hi Ralph,
This is a very complicated situation because not only is it a large multimer, but also one of the structures is a fusion protein with cytochrome b562.  

Certainly with ChimeraX's Modeller interface, you can do two types of calculations: comparative modeling (modeling based on templates) and filling in missing segments (modeling not based on templates, aka Model Loops).  Not sure you need to do any comparative modeling in this case, however, and I'm not clear on what is missing after you put the two structures together.

If you were just overlapping and then wanted to make a combined model from the non-overlapped parts, that is not really a Modeller thing.  You would just superimpose and then delete the redundant atoms and combine the rest into a single model.  You might want to delete the cytochrome b562 insertion as well.  The page for 7KOO at RCSB PDB tells me this fusion insertion is residues 353-457.  Unfortunately it's a huge mess to combine the two multimers because the chain IDs aren't arranged in the same way in the two structures.  These commands would open the two structures, keep only the first NMR model, delete the insertion, and superimpose (note the different orders of chain IDs in the specs):

open 7rpm
close #1.2-end
open 7koo
hide atoms
show ribbons
delete #2/A-E:353-457
mm #1/A,E,D,C,B to #2/A,B,C,D,E pair ss

So, if you were trying to make a single atomic model for further calculations, besides simply combining (see "combine" command), it would probably also involve a fair amount of manual text-editing of PDB files. (If this is just for a schematic figure, you could just hide the overlapping parts instead of trying to make a usable single model with all this combining, however.)

Say you have a model that combines these two existing structures.  Are there still parts that are missing?  Open the full UniProt sequence:

open uniprot:acha7_human
sequence associate /A-E acha7_human

If there is still something missing relative to the full UniProt sequence, you would use that sequence as the target for Model Loops.  With loop or missing-segment modeling using Modeller, there is no template.  It just starts with the structure you already have and generally keeps it the same, except fills in the missing parts relative to the target sequence.  However, this is only going to reliably model quite short segments (e.g. loops of <10 residues).  It's not going to predict the structure of a whole domain, for example.  If you don't have any template for the missing part, although that part will be built, it will probably look like a pile of spaghetti.

That made me think, well, what about AlphaFold prediction?  The feasibility of running a new calculation especially on a large multimer may be low  (and can take many hours if it even finishes) but the following commands (starting with an empty session) would get the single-chain prediction 5 times and match it to the chains in 7koo:

open 7koo
hide atoms
show ribbon
delete /A-E:353-457
delete /F-J
alphafold match #1/A-E trim false
hide #1 models

Before the alphafold fetch/match part, the above commands get rid of the fusion protein insertion and the bungarotoxin chains so that only the AchR chains are left.  There will be error messages that the cytochrome model structure could not be matched, which can be ignored since you don't care about a prediction for cytochrome b-whatever anyway.

AlphaFold is good at combining what is known from existing structures together, but it may or may not be able to predict the part without any similar known structures.  The red/orange parts are the parts deemed unreliable or disordered, and yellow is medium. See ChimeraX alphafold help:
<https://rbvi.ucsf.edu/chimerax/docs/user/commands/alphafold.html>

Best,
Elaine
-----
Elaine C. Meng, Ph.D.                       
UCSF Chimera(X) team
Department of Pharmaceutical Chemistry
University of California, San Francisco

> On Feb 14, 2022, at 1:44 PM, Ralph Loring via ChimeraX-users <chimerax-users at cgl.ucsf.edu> wrote:
> 
> Hi,
> The intracellular domain (ICD) of Alpha7 nicotinic receptors, like almost all eukaryotic Cys-loop receptors, is not resolved using X-ray crystallography or  cryo-EM.  Last week the PDB for an ensemble of NMR ICD structures for the receptor was released (PDB 7RPM).  I've been unsuccessfully trying to use ChimeraX and Modeller to make a model of a subunit that includes the ICD.  The 7RPM models include all 4 transmembrane domains and the ICD (but lacks the extracellular domain), so there is considerable overlap with the 7KOO PDB fo
> a7.  I started by deleting  all but one of the identical subunits in the 7KOO pdb and loaded one of the 7RPM models and tried to use Modeller to make a construct that includes both the extracellular domain and the ICD.  Unfortunately, even though both 7KOO and the single 7RPM are in the sequence alignments window, the ChimeraX-Modeller combination ignores the structure that is not chosen as the target (e.g. if 7RPM is the target, there is no extracellular domain, but if 7KOO is the target, there is either no ICD or the ICD that is shown bears no relevance to the structure in 7RPM).  I can align the two structures well with a low RMSD using Matchmaker because of the TMDs overlapping.  I just want to fill in the missing parts.  Is there a special trick to get Modeller to fill in the missing parts using ChimeraX as the interface? I'm not that experienced using Modeller.
> Thanks,
> Ralph H. Loring
> Associate Professor of Pharmacology
> Department of Pharmaceutical Sciences
> 166 TF
> Northeastern University
> 360 Huntington Avenue 
> Boston, MA 02115 USA
> 617-373-3216 office
> 617-373-8886 fax
> r.loring at northeastern.edu
> _______________________________________________
> ChimeraX-users mailing list
> ChimeraX-users at cgl.ucsf.edu
> Manage subscription:
> https://www.rbvi.ucsf.edu/mailman/listinfo/chimerax-users