Dear Tereza,
In certain cases it could be better to do a step back to be able to
rebuild properly.
Did you look carefully in the real-space the agreement between your
model and the visible density.
If you "over"refined your model in reciprocal space only you can loose
some information.
I comment you last comments :
1) Density too weak. Was it like that from the beginning, what happen if
you remove the "invisible" parts, is any differences green or red
density visible in the vicinity? Is that a flexible loop ?
The "systematic" coincidence between weak density and ramachandran
outlier is suspicious.
2) Manual NCS definition could help in a first approach but once your
model is complet enough it could be good to relax this constrain,
because you are actually fitting the model with what you think it should
be and what it actually is. In my opinion it's interesting to at least
try to use automatic which maybe will keep as NCS only the parts which
are really NCS. Sometimes very similar conformation of domain or
sub-units are close to be an NCS but not anymore NCS because already too
different (I am not sure to be clear ;-) ).
4) pdb redo even if really powerful and efficient can't be magic, if
your model suffer from mistake such as really bad rotamer or
ramachandran outlier, it could be impossible to revert that. Only a
manual intervention could fix it.
If the ramachandran outlier are in a weak density area, I would probably
try to fix that localy based on geometrical constraint (included
ramachandran) and then redo a round of refinement in reciprocal space to
see what happen. Sometimes the distortion is that important that you
need to re-build more than the residues directly involved especially at
relatively "low" resolution.
Hope this help.
Nicolas
Nicolas Foos
PhD
Structural Biology Group
European Synchrotron Radiation Facility (E.S.R.F)
71, avenue des Martyrs
CS 40220
38043 GRENOBLE Cedex 9
+33 (0)6 76 88 14 87
+33 (0)4 76 88 45 19
On 08/03/2019 10:36, Tereza Skalova wrote:
Thank you for your comments.
1) manual correction in Coot does not work - the density is too weak
2) manual NCS is substantially better than automatic local NCS in this
case
3) CPP4i2 might be good idea
4) PDB REDO is great, however no more help in this case
5) Prosmart - I use "prosmart -id -p1 target.pdb -p2 external.pdb" , I
will study other possibilities
Tereza
pá 8. 3. 2019 v 10:25 odesílatel Robert Nicholls
<nicho...@mrc-lmb.cam.ac.uk <mailto:nicho...@mrc-lmb.cam.ac.uk>> napsal:
Dear Tereza,
It is highly recommended that you do not attempt to directly
optimise the Ramachandran plot during refinement. Doing so would
not guarantee you a better model, and would mean that you could no
longer use the Ramachandran plot for validation purposes.
I suggest that you inspect each of the residues corresponding to
the outliers, and assess the conformation, geometry and density
fit of that residue and residues in the surrounding region. There
are various tools in Coot to help you with this. It should be
clear which regions are in need of attention (outliers that should
be fixed) and which "outliers" should be considered acceptable.
Indeed, ensuring that there are no Ramachandran outliers is not an
objective/requirement for a good model.
I refine in Refmac, using h-bond based Prosmart restraints based
on PDB structures (identical molecules with high resolution)
Prosmart h-bond based restraints, and Prosmart restraints based on
PDB structures are two different things. Are you using Prosmart
h-bond restraints, or Prosmart restraints to a high-resolution
homologous model? If you're using restraints to a high-resolution
homologue, are you generating restraints for all of your chains,
or just some of them? If you're just generating restraints for
some of them, then you should ensure that the others are
appropriately restrained also.
I use NCS, medium between AB (protein 1) and loose between CDE
(protein 2).
From your mention of "medium" and "loose" NCS restraints, I'm
guessing you're using CCP4i. Why not try using CCP4i2? This is the
currently recommended and supported interface for CCP4 software.
Don't use manual NCS restraints (medium, loose, etc.). Try using
automatically generated local NCS restraints - we find they work
better.
How are your R/Rfree behaving in refinement? If you're using
Prosmart restraints to homologous models then do you need/benefit
from the use of NCS restraints too, or are they working against
each other?
Best regards,
Rob
On 8 Mar 2019, at 09:08, Tereza Skalova <t.skalova.c...@gmail.com
<mailto:t.skalova.c...@gmail.com>> wrote:
Dear all,
I have structure at 3.3A resolution and I have ca. 35
Ramachandran outliers.
Do you have any idea how to reduce the number?
I refine in Refmac, using h-bond based Prosmart restraints based
on PDB structures (identical molecules with high resolution) and
I use NCS, medium between AB (protein 1) and loose between CDE
(protein 2). I use overall B-factor and 8 TLS groups.
Is it possible to optimize Ramachandran plot directly in Refmac?
Thank you
Tereza Skalova
------------------------------------------------------------------------
To unsubscribe from the CCP4BB list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=CCP4BB&A=1
------------------------------------------------------------------------
To unsubscribe from the CCP4BB list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=CCP4BB&A=1
########################################################################
To unsubscribe from the CCP4BB list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=CCP4BB&A=1
