From: Yuri Pompeu
Date: 30 September 2011 08:01
Hello everyone,
I am refining a structure to 2.4A with 2-fold NCS and twinned.
Maps look ok and Rfree is 0.27however as I start checking my validations I notice that after refinemnt
my geomtry gets significantly worse. especially the rama plot. Initially I have 2 outliers and I end up with 32 (5%)!!!
I played with the Xray weight term but alll it helped me with was rmsd bond/angles, rama is still messing up...
Can I impose some ramachadran restraints or maybe have a reference model?
best,
----------
From: Tim Gruene
Hello Yuri,
the ramachandran restraints can be imposed on with coot. I am not aware
if refmac5 does that directly, but using a reference model should fix it.
As far as I understand you cannot directly put in the reference model
but have to use a tool like prosmart by Rob Nicholls
(http://www.ysbl.york.ac.uk/mxstat/Rob/ - when I asked for a copy of the
program I received a reply within the day :-) )
which creates a text file with external restraints. The user
documentation of prosmart explains how to use it with refmac5, and it
works really well.
Cheers,
Tim
- --
Dr Tim Gruene
----------
From: Eleanor Dodson
Eleanor
----------
From: George M. Sheldrick
Dear Yuri,
Since the number of reflections depends inversely on the cube of the resolution,
it is easy to show that a perfectly twinned 2.4A structure will have the same
data to parameter ratio as an untwinned 3.0A structure. This determines how much
you can refine and which restraints are appropriate. In my experience, the 2.4A
twinned difference maps will be even less informative than 3.0A difference maps.
Fortunately, as is often the case with twinned structures, you also have 2-fold
NCS, so you can recover part of the damage by applying tight NCS restraints.
Best wishes, George
Prof. George M. Sheldrick FRS
Date: 30 September 2011 08:01
Hello everyone,
I am refining a structure to 2.4A with 2-fold NCS and twinned.
Maps look ok and Rfree is 0.27however as I start checking my validations I notice that after refinemnt
my geomtry gets significantly worse. especially the rama plot. Initially I have 2 outliers and I end up with 32 (5%)!!!
I played with the Xray weight term but alll it helped me with was rmsd bond/angles, rama is still messing up...
Can I impose some ramachadran restraints or maybe have a reference model?
best,
----------
From: Tim Gruene
Hello Yuri,
the ramachandran restraints can be imposed on with coot. I am not aware
if refmac5 does that directly, but using a reference model should fix it.
As far as I understand you cannot directly put in the reference model
but have to use a tool like prosmart by Rob Nicholls
(http://www.ysbl.york.ac.uk/mxstat/Rob/ - when I asked for a copy of the
program I received a reply within the day :-) )
which creates a text file with external restraints. The user
documentation of prosmart explains how to use it with refmac5, and it
works really well.
Cheers,
Tim
- --
- --
Dr Tim Gruene
----------
From: Eleanor Dodson
Dont foreget the usual reason for Rama outliers is model error! Check c=O are correct - cis peptides correctly defined in PDB ( coot and REFMAC USED to get these knotted.. ) etc etc..
Eleanor
----------
From: George M. Sheldrick
Dear Yuri,
Since the number of reflections depends inversely on the cube of the resolution,
it is easy to show that a perfectly twinned 2.4A structure will have the same
data to parameter ratio as an untwinned 3.0A structure. This determines how much
you can refine and which restraints are appropriate. In my experience, the 2.4A
twinned difference maps will be even less informative than 3.0A difference maps.
Fortunately, as is often the case with twinned structures, you also have 2-fold
NCS, so you can recover part of the damage by applying tight NCS restraints.
Best wishes, George
--
Prof. George M. Sheldrick FRS
No comments:
Post a Comment