Hi -
A quick note -
We (with Chris Ulens) have seen recently in one case that in a case
MATRIX AUTO would give 'excessive' RMSD, close to 0.030.
Checking the cell based on the projection of RMSDs to the cell axes,
a utility kindly provided by WHATCHECK,
it was suggested that the cell might be off. We changed the cell in
the PDB and MTZ and re-refined.
Now, AUTO gives lower RMSDs. If you think about it for a bit, it does
make sense.
The cells we get, especially since sometimes - and most
unfortunately - speed in processing is seen
as more important than being careful (what? you left the synchrotron
before you processed your data ! Shame!)
one or more cell parameters are often inaccurate and this can show is
RMSDs.
I am sure that now we can have a lively discussion about if we are
"allowed' to change the cell by 0.4 A based on
WHATCHECK output, or we need to re-process or re-measure the data ;-)
I am sure Chris can post numbers if requested - I suspect he reads
ccp4bb from his new headquarters in Belgium.
Tassos
On 13 Sep 2007, at 8:25, Jan Dohnalek wrote:
Just to add to the outliers in Refmac - careful when some of your
atoms have zero
occupancy. The bond restraints are not applied any more to these
atoms (my feeling) and
CERTAINLY are NOT CHECKED FOR OUTLIERS by Refmac at the end of the
refinement.
Jan Dohnalek
Eleanor Dodson wrote:
No comment on the Auto debate, but the RMSD of bonds is a crude
measure unless you check the outliers..
You can make REFMAC list these - I think it should do so by
default actually..
Bad metal distances for instance can distort the number enormously.
Eleanor
Mark Mayer wrote:
Hello all,
I believer the suggested matrix weight for running refmac is
'auto'. I'm at the end of refinment for a data set that I've cut
at 1.59 A, so maps are pretty. Here's the logfile with auto
weighting ...
NOTE bond rmsds !!!!!!!
REMARK 3 R VALUE (WORKING SET) : .15481
REMARK 3 FREE R VALUE : .18739
REMARK 3 BOND LENGTHS REFINED ATOMS (A): 4387 ; .
033 ; .022
REMARK 3 BOND ANGLES REFINED ATOMS (DEGREES): 5972 ;
2.776 ; 1.990
Am I correct that at this resolution one should aim for bond
rmsds of < 0.02. e.g for a weight of 1.0 the rmsd is 0.017, which
seems more reasonable.
REMARK 3 R VALUE (WORKING SET) : .16228
REMARK 3 FREE R VALUE : .19032
REMARK 3 BOND LENGTHS REFINED ATOMS (A): 4387 ; .
017 ; .022
REMARK 3 BOND ANGLES REFINED ATOMS (DEGREES): 5972 ;
1.689 ; 1.990
I know this has been discussed many times before, but expert
opinion always appreciated.
Thanks - Mark
