Detwinning data and refining against detwinned data is not the best
thing to do, especially when you realise that detwinning data with
twin fraction larger than 45% is done using model information in CNS:
the detwinned data you get out is not experimental anymore.
Subsequently, you cannot trust your R and Rfree statistics as you
would normally and any 'improvement' you see in your free R or your
maps, might be due to the 'detwinning'
Have you tried using TLS refinement in combination with twin refinement?
Peter
2007/6/9, Gregg Crichlow <[EMAIL PROTECTED]>:
Dear CCP4 Readers,
Quite some time ago, before the CCP4BB changed its server, I
posted the following inquiry concerning high B-factors in a protein-DNA
complex from a twinned crystal in which most of the DNA was not visible.
Dear CCP4BB:
We have a protein-DNA complex from a twinned crystal (space group P31). I
detwinned the data and refined the structure to 2.1 A with Rfree=28.8%
before adding water and nucleotides. We can only observe density for two of
the 25 nucleotides. I noticed that the B-factors are very high (around 50-60
A^2). Not many water molecules can be added using a 80 A^2 cutoff. We
determined the structure of the unbound protein, and the B-factors are
almost as high. I tried fixing the B-factors at 20, but Rfree increased to
30.9%. Is this indicative of an error that can be fixed? Thank you.
I do apologize for the delay in the follow-up. I thank everyone who
responded. The real problem was not the B-factors, of course, but the
incomplete DNA density. The various replies I received in essence mentioned
that (1) high B-factors are not necessarily a problem, (2) I should not
refine against de-twinned data, and (3) try density modification to be able
to see more of the structure. Unfortunately, the density modification
suggestion did not work with respect to improving the map, and refining
against twinned data had not appeared to be useful. It turned out that
de-twinning the data was the best way to proceed, judging by R/R-free
statistics, although this was not supposed to be as good as working with the
twinned data using the twinning fraction. I remember noting in the CNS
tutorial that using twinned data may not work well if the twinning fraction
is greater than 0.4, which it was in our case (about 0.49). Maybe that was
the problem. In the end, I used the data de-twinned in CNS as if it were a
perfect twin. Nonetheless, the high B-factors seem to be real. Although the
Wilson B-factor is high, maybe I was just hoping that this was due to the
DNA, not so much protein, and that sharpening the protein B-factors would
improve the map. This did not work. However, after handling the data the way
we did and now obtaining reasonable statistics, at least now I can be more
confident that the electron density for the nucleotides that we do see are
real. I thank everyone for their input.
Gregg
*******************************************
Gregg Crichlow
Dept. of Pharmacology
Yale University
P.O. Box 208066
New Haven, CT 06520-8066
*******************************************
