<div dir="ltr"><div>Hi Tom,</div><div>No, I was running only one subunit (502 amino acids) at a time. I'm trying other approaches with other software now to incorporate the ICD. When I bonded the different domains from multiple pdbs in Chimera, it really did turn into a Frankenstein. Also, I'd seen a youtube presentation suggesting that Modeller could handle more than one pdb, and I was hoping that would work as a way to combine the different domains. That didn't work out either. But as far as learning syntax from the log, that only works for commands that you can run from pull-down menus. All I know is that the syntax for the bond command in ChimeraX is different than the equivalent in Chimera. If someone could just give me an example of a script that picks the atoms and makes a CN peptide bond, that would be very thankful.</div><div>Thanks for all your help,</div><div>Ralph<br></div></div><br><div class="gmail_quote"><div dir="ltr" class="gmail_attr">On Fri, Feb 18, 2022 at 1:21 AM Tom Goddard <<a href="mailto:goddard@sonic.net" target="_blank">goddard@sonic.net</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div>Hi Ralph,<div><br></div><div> If you are trying to run alphafold to predict the acetylcholine receptor alpha7 pentamer, that is 2500 total amino acids (500 per alpha7 monmer), then there is no way that is going to run on Google Colab which fails beyond about 1000 amino acids. (Some alphafold size tests here: <a href="https://www.rbvi.ucsf.edu/chimerax/data/alphafold-jan2022/afspeed.html" target="_blank">https://www.rbvi.ucsf.edu/chimerax/data/alphafold-jan2022/afspeed.html</a>) Colab Pro definitely will not help. Also all Google Colab AlphaFold scripts including what ChimeraX uses run a limited alphafold with no structure templates. See the main AlphaFold Google Colab page for more details</div><div><br></div><div><a href="https://colab.research.google.com/github/deepmind/alphafold/blob/main/notebooks/AlphaFold.ipynb" target="_blank">https://colab.research.google.com/github/deepmind/alphafold/blob/main/notebooks/AlphaFold.ipynb</a></div><div><br></div><div>If you could install and run full AlphaFold on say an Nvidia RTX 3090 with 24 GB of GPU memory it might succeed (not run out of memory), but it is near the size limit. You would probably need a much more exotic Nvidia A40 (48 Gbytes) or A100 or similar. And even if you run that, you are expecting too much if you think it is going to give you a fully accurate prediction -- AlphaFold produces lots of poor models for large structures.</div><div><br></div><div> One more reality check on the capabilities of AlphaFold. It has no option to tell it exactly what PDB templates you want to use -- it chooses the "best 4", I have never seen the criteria documented for what "best 4" means, and further the log output from AlphaFold does not even tell you which 4 templates it uses. You could probably point it to a PDB database with only 7RPM if you wanted to just use that as a template if you can setup your own AlphaFold installation.</div><div><br></div><div> In summary, definitely don't expect miracles or ease of use from AlphaFold.</div><div><br></div><div> Of course the hand construction of Frankenstein models described by Elaine is likewise no easy task.</div><div><br></div><div><span style="white-space:pre-wrap"> </span>Tom<br><div><br><blockquote type="cite"><div>On Feb 17, 2022, at 3:48 PM, Ralph Loring via ChimeraX-users <<a href="mailto:chimerax-users@cgl.ucsf.edu" target="_blank">chimerax-users@cgl.ucsf.edu</a>> wrote:</div><br><div><div dir="ltr"><p class="MsoNormal" style="margin:0in 0in 8pt;line-height:107%;font-size:11pt;font-family:"Calibri",sans-serif">Hi Elaine Meng,<span></span></p><p class="MsoNormal" style="margin:0in 0in 8pt;line-height:107%;font-size:11pt;font-family:"Calibri",sans-serif">Thanks so much for your suggestions!<span> </span>However, I’m not thrilled about the Alphafold
idea, as I’ve tried to use it for receptors and keep getting booted out before
the job finishes (I’d like to get Colab Pro, but I can’t figure out how to sign
up with a university credit card; it’s only $0.55 per month in taxes, but my account
manager will go ballistic if I pay any taxes and Google doesn’t answer my emails
about how to pay with a tax exempt card).<span>
</span>Anyway, I got Alphafold to work for a 22 amino acid mu-conotoxin (<span>Mu-Conotoxin P0C349.pdb) and the predicted structure (in blue)
wasn’t great (see attached).<span> </span>The RMSD between
the Alphafold predicted pdb and the original (tan) was 4.1 angstroms, so not
good.<span> </span>If there is some way to guarantee
that Alphafold would use the 7RPM.pdb as a model for the intracellular loop, it
might work, but the human alpha7 prediction from Alphafold before 7RPM was released
(AF-P36544) is not good for the intracellular loop. <span> </span>There is now another Alphafold version (AF-</span>A0A1W2PN81)
that is equally bad.<span></span></p><p class="MsoNormal" style="margin:0in 0in 8pt;line-height:107%;font-size:11pt;font-family:"Calibri",sans-serif">Your scripts were incredibly helpful, as I haven’t figured
out the syntax for using the command line.<span>
</span>Is there a tutorial on using commands in ChimeraX? (I have the list of
commands in the user guide [<a href="https://www.cgl.ucsf.edu/chimerax/docs/user/index.html#commands" target="_blank">https://www.cgl.ucsf.edu/chimerax/docs/user/index.html#commands</a>],
but only rarely does that include examples of how the commands are used and I can't figure out the syntax code).<span> </span>What I’d like to do is to delete all the
overlapping amino acids in either 7EKI.pdb (which has eGFP in the cytoplasmic
loop instead of cytochrome) or 7KOO.pdb with one of the 7RPM ensemble and then
bond the ends together to get an approximation of what the intact receptor
would look like without overlaps.<span> </span>But I
can’t get the bond command correct in a test case using the two versions of
mu-contoxin (see attached csx file).<span> </span>I
know I want to bond #1Cys22CA to #2Arg1, but I don’t even know how to specify
the N-terminal nitrogen. (I can make the bond in Chimera using a combination of
Select/Atom Specifier and then Tools/Structure Editing/Build Structure/Join Models,
but that approach avoids using commands and I want to know how to do this in
ChimeraX). I know I’m going to have to futz with phi and psi after making the
bond, but I can’t even get that far (and there's two bonds to make).<span></span></p><p class="MsoNormal" style="margin:0in 0in 8pt;line-height:107%;font-size:11pt;font-family:"Calibri",sans-serif">By the way, Matchmaker does a good job of lining up the
overlapping amino acids with any combination of the receptor pdbs. <span> </span>As you can see, they are very comparable (the
dotted lines point down rather than up in your method). But I’d like to get a
pdb with a single chain for each complete subunit at some point.<span></span></p><p class="MsoNormal" style="margin:0in 0in 8pt;line-height:107%;font-size:11pt;font-family:"Calibri",sans-serif">Thanks for all your help! Any suggestions would be most appreciated.<br></p><p class="MsoNormal" style="margin:0in 0in 8pt;line-height:107%;font-size:11pt;font-family:"Calibri",sans-serif">Ralph Loring<br></p><div style="margin:0in 0in 8pt;line-height:107%;font-size:11pt;font-family:Calibri,sans-serif"><span></span><br></div><div style="margin:0in 0in 8pt;line-height:107%;font-size:11pt;font-family:Calibri,sans-serif"><span><span> </span></span><br></div>
</div><br><div class="gmail_quote"><div dir="ltr" class="gmail_attr">On Mon, Feb 14, 2022 at 6:33 PM Elaine Meng <<a href="mailto:meng@cgl.ucsf.edu" target="_blank">meng@cgl.ucsf.edu</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">To clarify, the "alphafold match" command would get already-made single-chain models from the freely available AlphaFold Database. It does not run a new AlphaFold calculation. It just superimposes the single-chain predictions onto the multimer structure that was already open, which is not the same as using it to predict a multimer. However, in practice the result is often quite reasonable, if there are already experimentally known structures with the same type of multimerization.<br>
<br>
I also meant to include more help links in the previous reply...<br>
<br>
matchmaker<br>
<<a href="https://rbvi.ucsf.edu/chimerax/docs/user/commands/matchmaker.html" rel="noreferrer" target="_blank">https://rbvi.ucsf.edu/chimerax/docs/user/commands/matchmaker.html</a>><br>
<br>
combine<br>
<<a href="https://rbvi.ucsf.edu/chimerax/docs/user/commands/combine.html" rel="noreferrer" target="_blank">https://rbvi.ucsf.edu/chimerax/docs/user/commands/combine.html</a>><br>
<br>
Model Loops<br>
<<a href="https://rbvi.ucsf.edu/chimerax/docs/user/tools/modelloops.html" rel="noreferrer" target="_blank">https://rbvi.ucsf.edu/chimerax/docs/user/tools/modelloops.html</a>><br>
<br>
Best,<br>
Elaine<br>
<br>
</blockquote></div>
<span id="gmail-m_-5602775228114855602gmail-m_3616561728441301583cid:f_kzrm5vdj0"><Superimposed AF- & native Mu-conotoxins.jpg></span><span id="gmail-m_-5602775228114855602gmail-m_3616561728441301583cid:f_kzrm7i0l1"><Bond AF-mu CTX to mu-CTX.cxs></span><span id="gmail-m_-5602775228114855602gmail-m_3616561728441301583cid:f_kzrm8v7r2"><Superimposed 7KOO without BGT and 7RPM Meng's method.jpg></span><span id="gmail-m_-5602775228114855602gmail-m_3616561728441301583cid:f_kzrm9k2t3"><Superimposed 7EKI and 7RPM.jpg></span><span id="gmail-m_-5602775228114855602gmail-m_3616561728441301583cid:f_kzrm9wnv4"><Superimposed 7KOO without BGT and 7RPM.jpg></span>_______________________________________________<br>ChimeraX-users mailing list<br><a href="mailto:ChimeraX-users@cgl.ucsf.edu" target="_blank">ChimeraX-users@cgl.ucsf.edu</a><br>Manage subscription:<br><a href="https://www.rbvi.ucsf.edu/mailman/listinfo/chimerax-users" target="_blank">https://www.rbvi.ucsf.edu/mailman/listinfo/chimerax-users</a><br></div></blockquote></div><br></div></div></blockquote></div>