Jack (?),
On Thu, 2010-12-09 at 21:35 -0800, Jack Russel wrote:
> So is it now the time to stop further refining the solution .
R-values are not the only criteria for this. You should be looking for
a) lack of unexplained density
b) good geometry
c) acceptable R-values
This
http://xray.bmc.uu.se/gerard/supmat/rfree2000/plotter.html
can help you to get some idea of the range of R-values for the
structures deposited in the PDB.
> if not then how can i lower the R factor and Rfree?
First, you should probably be looking to lower only Rfree. Otherwise,
the answer to your question depends on what you are dealing with. You
may have done some of this, but consider
- TLS
- more conservative geometry restraints
- NCS restraints if you got more than one copy in the asu
- there is this very cute Coot feature "Refine/improve Ramachandran
plot"
- etc, etc, etc
Keep in mind that, to paraphrase Daniel Gewirth, there are many ways in
which R-values can be manipulated, both deliberately and not. Yet
another reason not to go nuts about R-values and to take them, like
poison, with a grain of salt.
> The second question is it possible to have such a large difference
> between R factor and R free?
Well, apparently it is, you have just proven it :). Seriously though,
the 10% that you have is on the high side. Some things mentioned above
for the reduction of Rfree may help to close the gap, but once again,
stuff happens.
I'd make sure to have 10% set aside for the test set, at this resolution
5% may be not enough.
Could you privately send me the mtz-file from refmac output? I am
curious about R-value statistics in this case. Naturally, it will be
treated confidentially.
Cheers,
Ed.
--
"I'd jump in myself, if I weren't so good at whistling."
Julian, King of Lemurs
|