......and I may add that for DMMULTI two "rather nonisomorphous" data
sets of the same crystal form can work like a dream, in particular at
low resolution where solvent flattening and histogram matching can be
tricky, see for example Morth et al. 2007, Nature 450, 1043 & Pedersen
et al. 2007, Nature 450, 1111
Poul
On 15/04/2009, at 12.02, Pietro Roversi wrote:
Dear James and Andy,
my twopenny worth. I am
using dm and dmmulti to solvent
flatten after molecular replacement phases when no shred of
experimental
phases is available. If the model is severely incomplete
(say at least 25% missing or more) I use phases from Buster-TNT
refinement with missing atoms modelling switched on (this reduces
the model bias
imp[roves the scaling and helps bringing up weaker features from the
yet unmodelled bits).
I also blur the phase distribution either by multiplying the FOMs
by a number less than one (say 0.6-0.8) or by doing the same to the
Hendrickson-Lattmann
coefficients (the idea is driectly inspired from this paper:
Perrakis, A., Harkiolaki, M., Wilson, K. S. & Lamzin, V. S. (2001).
Acta Crystallogr D Biol Crystallogr 57, 1445-1450.
I tend to try a few values of the multiplicative factor
and choose the one that gives the maps that I like the best.
Ah: it may be trivial again, but for the project I am working on at
the moment even a 6.7 A dataset
in a different crystal form has made a big difference! the dmmulti
fourfold averaging across
2 crystal forms being much better than the twofold averaging in dm
in the high-resolution
form. Admittedly this extra crystal form has 90% solvent! So if
possible at all
use dmmulti and multi-crystal averaging.
I hope this helps!
Pietro
--
Pietro Roversi
EP Abraham Fellow in Biochemistry - Lincoln College - Oxford
Sir William Dunn School of Pathology, Oxford University
South Parks Road, Oxford OX1 3RE, England UK
Tel. 0044-1865-275385
