Hello Natalia,
> I am working on a structure where half was built automatically by phenix and the other half, that has poor density, I built in Coot.
OK, so first things first. Only use Coot when automatic building
fails. Try Buccaneer (as Eleanor says).
> Overall the part that I built makes total sense (secondary structure, loops, etc), but positions of many of the side chains and even some of CAs are only approximate.
Yes, not unusual.
> What I used to do in the beginning is to build a part (say add around 10 residues)
OK,
> and then run real space refinement for the whole molecule in Coot,
Non optimal, I'd say. I wouldn't recommend it.
> right now I came to conclusion that sometimes it results in worse model than the initial.
Yes, that can happen.
> So my question is what is the correct strategy for improving this kind of structure?
No two ways about it - building in low resolution maps is hard.
> Should I mutate all my side chains to alanine,
No. It's sometimes a good idea to "stub" a residue (delete the
sidechain) but almost never to convert to alanine.
> refine the structure in phenix and then proceed in Coot?
That's one way... Refmac + ProSMART is another.
> Or there is a way to move atoms by hand?
Yes, you can do that - but generally speaking you want to minimize the
amount that you have to do it.
Rotate/Translate Zone or Ctrl-{Arrow-keys} and Ctrl-Shift-{Arrow-keys}
to translate and rotate the active residue. If you really want to move
just one atom, you can activate the residue for Rotate/Translate and
Ctrl-LM-Drag on an atom.
> I mean is there a way to drag a certain atom while keeping its connections to other atoms, without loosing proper bond lengths, etc?
If you want to move whole fragments, or rotate about rotatable bonds,
the yes.
If you mean in refinement, then no (not yet). Having said that, Coot's
refinement is quite good at restoring distorted bonds to reasonable lengths.
Jan answered the NCS part.
Paul.
|