[chimerax-users] Refine predicted structure by pulling

Mon Oct 25 13:37:13 PDT 2021

Result of some quite aggressive stretching (alternating pulling on residues 2000 onwards and 1-700, extending the tugged selection at both ends as things straightened out). This was with my dev ISOLDE build allowing me to use the "adjustForConfidence true" argument and an 8 Angstrom distance cutoff, which definitely helped keep things under control (also ended up restraining torsions with "isolde restrain torsions #1". Quick-and-dirty video overview of the final model at https://drive.google.com/file/d/15zCt7WF7W6rBLoujMhFwgdpDYaPsYMB-/view?usp=sharing, showing that almost all the distance restraints remain well satisfied, except for a few at the hinges.

Hope you find it useful!

Tristan

________________________________
From: ChimeraX-users <chimerax-users-bounces at cgl.ucsf.edu> on behalf of Tristan Croll via ChimeraX-users <chimerax-users at cgl.ucsf.edu>
Sent: 25 October 2021 21:14
To: goddard <goddard at sonic.net>; chimerax-users at cgl.ucsf.edu <chimerax-users at cgl.ucsf.edu>; Roden Deng Luo <deng.luo at kaust.edu.sa>
Subject: Re: [chimerax-users] Refine predicted structure by pulling

Hi Tom,

Yes, "kappa" is the stiffness. Default in the current release is 5, (as specified in the documentation via "usage isolde restrain") which I agree is very soft for this particular purpose - after playing around with it a bit I find "isolde restrain distances #1 distance 5 kappa 100 fallOff -3" works quite well. The extra "adjustForConfidence true" argument coming in the new ISOLDE release (not yet available - here in a few weeks) will help further - because it uses AlphaFold's own confidence on residue-residue distances, it can extend the distance restraints over a longer range (e.g. the default 8 Angstroms) while ignoring spurious chance contacts between residues. Will play around a little more and share the result.

-- Tristan
________________________________
From: Tom Goddard <goddard at sonic.net>
Sent: 25 October 2021 20:01
To: chimerax-users at cgl.ucsf.edu <chimerax-users at cgl.ucsf.edu>; Roden Deng Luo <deng.luo at kaust.edu.sa>; Tristan Croll <tic20 at cam.ac.uk>
Subject: Re: [chimerax-users] Refine predicted structure by pulling

Hi Roden, Tristan,

  I tried Tristan's suggestion for straightening the long AlphaFold helix model.  It didn't work well for me because the restraints were too soft, so if I pulled on the first 160 residues it hardly moved the attached 2000 and instead just distorted the linker.  Tristan is there a way to make the restraints much stiffer?  I see the "isolde restrain distance" command has a "kappa" option and "falloff" option but I don't know if kappa is a stiffness or what magnitude it usually uses.

  I think a better approach is to just rigidly rotate and translate large straight sections by hand using the "Move atoms" mouse mode in the ChimeraX 1.3 toolbar tab called Right Mouse.  I tried this and have attached an image of about 10 minutes fiddling by hand.  I hover the mouse at a hinge residue to get the residue number and then select everything on one side of that hinge "select :1182-2419" then use the right mouse button in the "move atoms" mode, holding the shift key down to do rotations, and no shift key to do translations.  I zoom in on the hinge to try to adjust so the bond is not stretched and to avoid clashes.  It would probably take 1 hour to do a good job.  With my 10 minute test I then saved an mmCIF file and tried in ChimeraX 1.2.5 to minimize it with ISOLDE, but it said there were severe clashes.

    Tom

On 10/24/2021 1:49 AM, Tristan Croll via ChimeraX-users wrote:
Hi Roden,

This should be reasonably straightforward with the ISOLDE plugin (https://cxtoolshed.rbvi.ucsf.edu/apps/chimeraxisolde). You'll also want to be working on a machine with a good GPU to make it feasible for a model this size. It'll be easier still with the new release coming in a few weeks, which has some added tools specifically for working with AlphaFold models. But with care you could do it now:

  *   first, check carefully for entanglement between distant domains (this is an issue for AlphaFold with these very long, flexible proteins - if distant parts of the protein have no coevolutionary signal they don't really "know" about each other and can end up trying to occupy the same space). Looks like your particular example is free of that problem.
  *   Start ISOLDE. Then, assuming your protein is model #1, do the following commands:
     *   isolde restrain distances #1 distanceCutoff 5
     *   isolde restrain torsions #1 angleRange 60
     *   isolde sim start #1
  *   Then, you'll want to use the "tug selection" mouse mode (third from bottom right of the ISOLDE panel). Make a selection for a suitable region to tug (e.g. "sel #1-100 at CA" to tug the N-terminus), and tugging forces applied by right-click-and-drag will be spread over that selection. Keep a close eye on the distance restraints - if they start to become too strained (lots of purple appearing), stop tugging and let the model settle.
  *   If you have unwanted distance restraints between domains that should be distant from each other, you can selectively release restraints on selected atoms with "isolde release distances sel".
  *   You might find it easier for a job like this to reduce the view to a C-alpha trace - just "hide ~@CA" - if you do this after you've started a simulation, it'll automatically revert to the all-atom view once you stop.

Happy modelling!

-- Tristan

________________________________
From: ChimeraX-users <chimerax-users-bounces at cgl.ucsf.edu><mailto:chimerax-users-bounces at cgl.ucsf.edu> on behalf of Roden Deng Luo via ChimeraX-users <chimerax-users at cgl.ucsf.edu><mailto:chimerax-users at cgl.ucsf.edu>
Sent: 23 October 2021 16:30
To: chimerax-users at cgl.ucsf.edu<mailto:chimerax-users at cgl.ucsf.edu> <chimerax-users at cgl.ucsf.edu><mailto:chimerax-users at cgl.ucsf.edu>
Subject: [chimerax-users] Refine predicted structure by pulling

Dear ChimeraX Users,

There is this AlphaFold prediction: https://alphafold.ebi.ac.uk/entry/P02549. And we know this protein should be more or less in a rod-like shape rather than a globular shape: https://en.wikipedia.org/wiki/Spectrin and related literature such as https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3218374/. I wonder is there a way that we can pull the two ends of the predicted structure while maintaining the secondary structures, mostly alpha-helices in this case, inside ChimeraX?

I am quite new to ChimeraX. I understand this is more fitted to an MD simulation. But I got to know there are options to fit a PDB to a density map. So I wondered if this can be achieved in ChimeraX.

Best,
Roden

________________________________
This message and its contents, including attachments are intended solely for the original recipient. If you are not the intended recipient or have received this message in error, please notify me immediately and delete this message from your computer system. Any unauthorized use or distribution is prohibited. Please consider the environment before printing this email.

_______________________________________________
ChimeraX-users mailing list
ChimeraX-users at cgl.ucsf.edu<mailto:ChimeraX-users at cgl.ucsf.edu>
Manage subscription:
https://www.rbvi.ucsf.edu/mailman/listinfo/chimerax-users

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://plato.cgl.ucsf.edu/pipermail/chimerax-users/attachments/20211025/785b2b78/attachment-0001.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: spectrin_extended.zip
Type: application/x-zip-compressed
Size: 609177 bytes
Desc: spectrin_extended.zip
URL: <http://plato.cgl.ucsf.edu/pipermail/chimerax-users/attachments/20211025/785b2b78/attachment-0001.bin>