[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[ccp4bb]: Symmetry breakage, twinning, high b's ...
*** For details on how to be removed from this list visit the ***
*** CCP4 home page http://www.ccp4.ac.uk ***
Dear Protein Crystallography Experts,
I am refining a structure of a biological (homo)dimer which should, in
principle, have one AMP ligand in each subunit. Data (up to 2.2 A, good
completeness and Rmerge) could be processed ,
say, reasonably well, in P6122. The PDB contains one homologue structure
(86% identity, 92 % homology, same space group), which gave a clear AMoRe
solution, practically in the same position. In the mr model, the biological
2 fold is a crystallographic one, such that I have, in P6122, one monomer /
au. I carried out the refinement, which got stuck at:
"Overall R factor = 0.2258 Free R factor = 0.2873 "
also, bad features are that average B's are ~41 A^2 (so high, could
them be
telling something ?) and the model has several regions where the fitting of
the model to the density is not so convincingly (needs 1 sigma, for some
main chain regions, even out of loops), but specially through the AMP
binding site and in this last one itself, which has a breakage in the
density through it.
Besides all that, anisotropy could be observed in the intensity
distribution (only between h even and odd and between k even and odd,
systematically higher when they are odd, in all resolution ranges), truncate
output :
" For each class, Mn(I/sig(I)) is given for even and odd parity with respect
to the condition,
eg group 1: h even & odd; group 7 h+k+l even & odd; group 8 h+k=2n & h+l=2n
& k+l=2n or not
Range Min_S Dmax Nref 1 2 3 4 5
6 7 8
h k l h+k h+l
k+l h+k+l h+k,
1 0.00111 14.59 67 51.4 57.0 48.9 60.5 53.3 54.5 54.4 53.4 54.2 53.6
54.0 53.8 55.0 52.9 54.8 53.6
2 0.00470 10.99 91 59.4 69.5 58.2 75.2 62.0 66.8 63.2 65.7 63.6 65.2
63.7 65.1 65.5 63.3 61.7 65.3
3 0.00829 9.18 111 59.2 64.6 57.3 66.4 61.8 60.6 64.9 58.2 60.4 62.1
61.3 61.1 59.3 63.2 63.8 60.4
4 0.01187 8.04 113 59.4 70.0 59.8 71.5 65.6 64.8 65.4 65.0 63.7 66.6
66.3 64.1 65.3 65.2 65.0 65.3
5 0.01546 7.25 141 58.3 62.8 53.9 68.8 60.4 59.8 60.6 59.7 60.6 59.6
59.9 60.4 58.1 62.2 60.8 59.9
6 0.01905 6.65 147 55.7 61.3 55.0 64.2 57.7 59.5 55.7 61.2 59.5 57.7
58.5 58.7 57.5 59.7 56.7 59.2
..... snipped ...
25 0.08720 3.32 282 35.4 37.4 36.5 36.4 35.6 37.2 35.9 37.0 37.6 35.2
36.0 36.8 36.4 36.5 36.6 36.4
26 0.09079 3.26 268 36.6 37.8 34.6 40.3 37.4 37.1 36.7 37.8 34.8 39.7
35.5 38.9 35.1 39.5 32.4 38.8
27 0.09437 3.20 273 33.6 35.3 32.2 37.2 35.2 33.6 34.4 34.4 34.7 34.1
32.7 36.1 34.6 34.2 32.9 34.9
28 0.09796 3.14 287 31.2 34.6 31.6 33.7 32.9 32.4 32.2 33.1 32.8 32.5
32.9 32.4 32.5 32.8 32.7 32.6
29 0.10155 3.08 270 26.0 30.4 25.9 30.5 29.2 26.6 26.3 29.6 28.7 27.1
29.5 26.3 28.3 27.5 28.6 27.7
"
So I suspected of symmetry breakage and/or twining.
I reprocessed the data in P61 (even at this sg I have a good
redundancy), and of course the above mentioned
anisotropy in intensity distribution disapeared. I also doubled my model,
such that now I have a dimer, subunits more or less related by the old
(supposed) crystallographic 2-fold (but not restrained one to the other in
the refinement). Yes, the R's did improve:
"Overall R factor= 0.2067 Free R factor = 0.2746"
Please, note that I kept the SAME Rfree set, i.e., when breaking the
symmetry, I took care to double the number of reflections in the Rfree set,
but this doubling was governed the the symmetry which I was breaking, that
is, I tried to preserve as much as possible the Rfree set.
But, the average B's keep ~ 41 A^2 !!!. On the other hand, at looking at
the ligand densities, one monomer shows very good connectivity for the
ligand
(quite good fit) and the other is bad (a little worse than when I was
working with only one monomer, sg. P6122). I run lsqkab, but did not see
expressive diference in the atom positions (even in visually
comparing, well the most distant (some few) ones ~ 0.2 A). And also, some
regions kept with the bad fitting between the
model and the electron density.
So, now, questions:
1) Could only a difference in the bound ligand be enough to lead to all the
problems I had when refining the structure in P6122 ? Could this lead to a
symmetry breakage such that I must refine the structure in a lower symetry
space group (P61) ?
2) Should I make a simulated annealing of my doubled model in P61 ? Could
this improve the density fitting ? Could my doubled model be biased by the
anterior model in P6122 ?
3) In the case of P61, at using two monomers, I double the number of
parameters. Of course this
should allow an artificial lowering of the R factor, but I also doubled the
number of reflections (hum...., although they are quite similar pairwise).
Who weights more here ? Doubled parameters, doubled data ? Should the R
artificially lower, raise or keep the same ? But, in the case where the data
are similar pairwise, should this not be decisive for artificially lowering
the R ?
4) In spite of the artificialism talked in question 2, would not the fact
that the Rfree lowered be significative (remember that I doubled the Rfree
set keeping the symmetry related ones by the removed old symmetry) ? How
reliable could this Rfree be to point me that there really is a symmetry
breakage ?
5) I heard that AMP is quite unstable ? What would be the product of its
degradation ? Maybe this could this be fit in the "broken" density.
6) Could I have a kind of "internal twinning" (another better or a correct
name for this ?) , such that the monomer which
preserves the intact AMP preferably (but not all of them) assumed one
position
at forming the crystal ? Although the ligand densities in both monomers are
decidedly different, they still have a slight resemblance...
Any suggestion for overcoming this problem at crystallizing the protein ?
7) Another question which I think is not directly related to the other ones.
Later on, I noticed the the structure factors of the protein used for the
molecurar replacement were available at the PDB (but not indicating the
R_free set). Since the high homology (92
%) and identity (85 %), similar (if not the same) space group and
orientation in the unit cell, should I use the same R-free set that the
mr model (if this one was flagged in PDB) ? What should be the criteria to
lead one to use the same set
(hum, I guess this is a very interesting discussion) ?
Any help, point to literature, etc..., appreciated.
Many thanks (and sorry for the rather long post),
************************************************************************
Prof. Jorge Iulek, Ph. D.
Protein Purification and 3D Structure Determination Group
Department of Chemistry
State University of Ponta Grossa - PR
Brazil
Office phone: ++ 55 (42) 220-3062
Home phone : ++ 55 (42) 224-6182
Fax : ++ 55 (42) 220-3342
e-mail: iulek@interponta.com.br
iulek@uepg.br
*****************************************************************