By Ana Casañal & Paul Emsley

This tutorial is designed for 0.9.1 or later.

Aim: we will fit a domain of the Cleavage and Polyadenylation Factor

1: Setting Up

1.1 Fetch the Files

Using a web browser, download the bundle for EMD-3908

Move that tar file here and extract it.

Let’s download the sequence of a fragment of the structure:

Move that sequence file to the current directory (for easy access).

Start Coot:

$ coot --map EMD-3908/map/emd_3908.map

1.2 Map Manipulation

Let’s see more of the map

  • Edit → Map Parameters → Map Radius EM → 70
  • OK

Let’s use a smoother map

  • Calculate → Modules → Cryo-EM
  • Cryo-EM → Sharpen/Blur…
  • activate the “Resample” checkbutton
  • Make Map
  • Close

You should now have an extra map (“emd_3908.map Blur 20.00”). Compare this maps with the original. You should find that the new (smooth) map is easier to understand.

  • use the Display Manager to delete the first (mrc) map.

Change the contour Step for the new Maps:

Display Manager → Properties → Change by rmsd 0.33 → OK

As a rule of thumb, a good default/starting contour level is 5.5 rmsd.

1.3 Get the Homolog:

  • File → Open Coordinates… coot_tutorial_2/6f9n.pdb
  • if Coot gives you a “Nomenclature Errors” dialog, just click “Cancel”
  • Use the Display Manager to change the representation
  • Instead of “Bonds (Colour by Atom)” use “C-alphas/Backbone”

Move back to the middle of the molecule map

  • Cryo-EM → Go To Map Molecule Middle

  • Now move the homolog to here (the centre of the map):
  • Calculate → Move Molecule Here
  • Choose the “6f9n” molecule
  • Move It

You can now change the colo(u)rs if you wish:

  • Display Manager → Properties → Colour
  • Edit → Bond Colours [slide the slider]

I like to work with a blue map and the model as orange or green (this is good colour scheme for working but not good for screenshots used in presentations).

2: Jiggle

  • Morph → Jiggle Fit This Molecule with Fourier Filter

It should roughly fit now. If it doesn’t, try jiggling again once or twice more.

3: Extract Our Fragment

Extract the worst-fitting (WD40) domain:

  • You can see that there are 2 chains.
  • We want to extract a wD40 domain from the larger A chain
  • Using Jones’ Rainbow, find the domain start and end residues numbers (you are trying to find a doughnut-shaped molecule that fits this doughnut-shaped density)
  • Let’s imagine that you think that the residues at the ends of the domain are 517 and 1011:
  • Edit → Copy Fragment → [Use 6f9n molecule:] //A/517-1011 → OK

Let’s work on this fragment:

  • Display Manager → Last Only
  • Reset View

Let’s delete the sidechains from the atom selection molecule (number 4 usually):

  • Calculate → Modelling → Delete Side-chains for Active Chain

For the most recent model (bottom of the list), in the Display Manager use

  • C-alphas/Backbone
  • OK [Dismiss the Display Manager]

4: Setup Refinement

We need to adjust the weighting of the map to the model:

  • R/RC → Refinement Weight → 60.00 → OK (don’t click “Estimate” we need tighter weights at the moment)

    Let’s add some local distance restraints:

  • Calculate → Modules → Restraints

    Usually 5.0A works well for models with no sidechains:

  • Restaints → Generate All Molecule Self Restraints 5.0
  • {Coot displays new self-restraints as thin grey lines} Briefly look at these restraints, then undisplay them:

  • Restaints → Undisplay Extra Distance Restraints

    Let’s tighten up the Geman-McClure restraints a bit:

  • Refine → Set Geman-McClure alpha 0.01

5: Refinement

This is where the power of Coot 0.9 becomes apparent.

{I’d like to note here that on a modern computer with multiple cores and threads, this is a pleasant experience. If that is not like your computer then the following section can be confusing and frustrating}

  • Refine → Chain Refine
  • {wait and watch, you can turn the view if you wish (be careful not to click an atom)… this takes about 20 seconds on my computer (a PC from 2019)}
  • When the refinement dialog says “Success”:
  • Accept Now let’s refine again with less tight Geman-McClure restraints:
  • Refine → Set Geman-McClure alpha 0.1
  • Refine → Chain Refine
  • {wait and watch ~10 seconds}

{A note on “yanking”: by which I mean “smooth displacement” - not jerky shifts}

When the refinement dialog says “Success,” this time we don’t yet press Accept… Examine the model, being careful not to inadvertently pull on an atom. Maybe you will see that there is a region that doesn’t fit, if so, yank on the worst fitting CA and pull it to where you think it should go.

  • Double-clicking on an atom will release the pull restraint

When you are happy, dismiss the Refinement dialog:

  • Accept

5.1 Redo

It can be non-trivial to decide what needs to move and how to move it. It is worth undoing your modifications and refining again for practice.

This time perhaps with the distance restraints shown:

  • Undo (left-facing arrow icon)
  • Calculate → Scripting → Python
  • set_draw_moving_atoms_restraints(1)
  • Refine → Chain Refine
  • yank as needed
  • OK

… or with different cut-off for the Geman-McClure restraints, or a different alpha for the Geman-McClure restraints, or a different weight for the map. Or a different blur for the map. You can delete the current extra restraints with Restraints → Delete All Extra Restraints.

  • Try proportial editing: with the Real Space Refinement active, use Ctrl Middle-mouse scroll to change the radius of the atoms affected by the atom pull displacement.

  • Test, play, refine, yank until satisfied.

Reset Geman-McClure alpha to 1:

  • Refine → Set Geman-McClure alpha 1

6: Review and Trim

Upon review, you will notice that there are parts of the model that don’t fit the map. Try yanking them around with Tandem Refine. Other parts of the model don’t have density - so delete the residue range - this may help the alignment we are about to do.

Maybe the density fit validation dialog will be useful? You might need to reset the weight: 1.5 seems like a good number

  • Validate → Density Fit Analysis [Choose the “atom selection from pdb6f9n” molecule]

7: Mutate

This is Coot’s version of “Homology Modelling” - except that the model geometry optimization occurs in the context of the experiemental data:

  • We have finished with the “Self Restraints” - let’s delete them:
  • Restraints → Delete Extra Restraints
  • Coot displays a molecule chooser dialog:
  • Select the “atom selection from pdb6f9n.ent” molecule

  • OK

  • Calculate → Use ClustalW for Alignment, Then Mutate

    The chain for mutation is the “atom selection from pdb6f9n.ent” and the target sequence is in the file CPF-X-domain.seq

  • {wait}
  • Refine → Chain Refine
  • {wait}
  • OK

8: Check & Edit

Go through the structure residue by residue looking for things to fix (this can take some time, so you might just do a few, for a flavour, before moving onto Section 8.2)

8.1 Do some Fix-ups

  • Out of register errors
  • Missing side-chains
  • Missing loops
  • Loops with too many residues in the model
  • Bad rotamers
  • Bad peptides
  • Clashes
  • Use: Validate → Alignment vs. PIR… to help you adjust the residue numbers

There will be places where you need to close (or open) a loop by renumbering residues.

  • Use Validate → Validation Outliers

    or

  • Validate → Overlaps, Peptides, CBeta, Rama and Rota Outliers

Side chains atoms can be added the residue at the centre of the screen using “K” and deleted using “Shift K”.

8.2 Unclash

Part of the work on the refinement has been to change the way non-bonded contacts are minimized. To reduce/remove atom overlaps (or “Clashes” as Molprobity would call them)

  • First determine the model number of the domain we are fixing - for me it’s 2
  • and now add hydrogen atoms to that model:
  • Calculate → Scripting → Python
  • coot_reduce(2)
  • Refine → Chain Refine
  • {wait - this takes a lot longer than before - there are lots of interactions now}
  • Accept

{For your notes, don’t act on this now, to delete the hydrogen atoms, use the Python scripting function delete_hydrogens(4) (or whatever your model number is)}

8.3 Ramachandran Outliers

It can be convenient to use Ramachandran Restraints to improve the model validation. This is how I do it:

  • Click the R/RC button
  • Tick Use Torsion Restraints
  • Tick Ramachandran Restraints
  • OK
  • Validate → Ramachandran Plot [and chose the “atom selection” molecule]
  • Outliers Only

Now click on a red spot - they may be many of them (maybe 50 or so):

  • Look at the model. Sometimes the outlier is problematic because there is a model-building error - you should not attempt to fix such problems using Ramachandran restraints.
  • But sometimes, things are good, it’s just that the main-chain torsion angles need tweaking:
  • Tandem Refine
  • Now flipping and back-flipping on This Peptide or the Next Peptide often fixes the problem.
  • If needed you can change the Ramachandran restraints weight using the menu items in the Refine menu. (But note however, that Coot refinement often gets upset when the Ramachandran weights are high.)

9: Done

You can check how well you did by comparing against the reference model - the accession code for that is 6oej coot_tutorial_2/6eoj.pdb