[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[ccp4bb]: Summary of AMore question
*** For details on how to be removed from this list visit the ***
*** CCP4 home page http://www.ccp4.ac.uk ***
Dear all,
Many thanks to these people who responded to my Amore question.
Here is a summary of their suggestions,
> My questions are
>
> 1.Could someone know why such CC and R-f doesn't produce right
> solution?
>
> 2. Does Amore check the molecular packing during translation search? How
> to implement it?
>
1. Dr. Frank Vondelft <Frank.Vondelft@syrrx.com>
Well, these rotations are all searched independently, and if your
spacegroup (which you don't give) has multiple origins, amore will place
each solution at some random origin.
What you should do is take the first solution ONLY and fix it while
finding the second molecule -- i.e. "locked" rotation search. Then, fix
BOTH these for while finding the third solution, and then fix these
three while searching for the fourth. With peaks that clear, they
should all fall out very easily.
2. Dr. Lijun Liu <liu@kuchem.kyoto-u.ac.jp>
>1.Could someone know why such CC and R-f doesn't produce right
>solution?
-------------------------------
You might not get the right solutions. A important thing is whether the CCs
and Rs have apparent gap with the 5th, 6th... solutions. If so, you got the
right ones; if no, you did not. In fact, this rule should also be applied
when you fixed the 1st solution for the 2nd one, then 2 for the 3rd, and
so on...
>2. Does Amore check the molecular packing during translation search? How
>to implement it?
-------------------------------
Of course, but it may depend on the CCP4 version. Please check the log file.
The last value of a solution string (alpah, beta, gamma, Tx, Ty, Tz, CC, R,
CCi, and the last value DMin), the DMins correspond to the center-center
distance estimated from the molecular packing! Apparently, you can judge
your solutions when estimated the size from the model. If this value is too
small, it may mean a wrong solution!
3. Dr. Ilana Agmon <chilana@techunix.technion.ac.il>
> But, problem appears when I check their packing. Molecules overlap to each
> other, either directly or to their symmetry mates.
Well, you spotted it correctly . What you got is 4 symm related positions
for one of your molecules.
You have now to fix this solution, which seems correct according to the
scores, and look for the other molecules in the AU.
> 2. Does Amore check the molecular packing during translation search? How
> to implement it?
As far as I know the CCP4 version of amore does not check the molecular
packing. However, the standalone version of Navaza allows you to set a
minimal COM to COM distance and ignore impossible solutions.
4. Dr. Eleanor J. Dodson <ccp4@ysbl.york.ac.uk>
Many reasons.. CC and R-f are indicators only.
Have you got the right space group? Is there translational non-cryst
symmetry? Look at the native 4A Patterson and see if there is a large
peak off the origin.. With 4 molecules in the asymm unit it is common to
find a peak with on coordinate 0.5. And that will generate apparent
absences for a 21 screw although your space group may well have not got
one.
You need then to run the TRAFUN in all possible space groups.
eg P 21 21 21 and P 2 21 21 if your NCS gave you a peak at x=0.5, ...
Your solution would appear OK because you may well have 2 molecules
correctly placed in the wrong space group..
If that is true you will probably have found a solution which is half
right -
5. Dr. Fred. Vellieux <vellieux@ibs.fr>
For difficult problems (i.e. multi-body problems) I find it easier to use
the molecular replacement calculations in CNS. 1 or 2-body problems are OK
with AMORE though.
An example is the 3-D structure of trypanosomal PPDK (the manuscript has
been submitted, I am expecting comments by referees). This was a 4-body
problem. Solved using CNS not AMORE (far too automatic for that).
6. Dr. Anastassis Perrakis <perrakis@nki.nl>
> 1.Could someone know why such CC and R-f doesn't produce right
> solution?
very common.
>
> 2. Does Amore check the molecular packing during translation search?
no.
> How to implement it?
with great difficulty.
> I really appreciate it if someone can share their valuable experience on
> this kind problem.
the following might sound trivial, but :
Did you use the FIRST rotation solution for translation functions after
fixing the first solution (which seems right ?).
In general, the second solution shoudl drammatically improve R factor.
So, I do believe the first solution but NOT the other ones.
Anyway, I am usually to bored to think and I run MOLREP....
easy anought to try from ccp4i ...
7. Dr. Juergen Bosch <jbosch@biochem.mpg.de>
Alternatively you could give Molrep a try with the option _PST Y or _PST C
http://www.ysbl.york.ac.uk/~alexei/molrep.html
8. Dr. Deena Oren <oren@cabm.rutgers.edu>
The solution does seem correct. Did you look at the maps? I would strongly
suggest doing solvent flipping and analyzing the maps themselves. It is
possible that the regions that overlap are actually different than the
model. Solvent flipping followed by simulated annealling had moved whole
loops into place for me under similar conditions where I had six molecules
(actually I used two trimers to solve the structure) in the a.u. and Rfact
of 48.5 with correlation of 22. The model was a pruned protein (took all
residues in common with a 20% homolog and the rest were Ala or Gly).
I have found twice now that the only way to convince yourself of the
correctness of the solution is to look at maps. They are, afterall, our
real data.
Thanks again for your valuable advice.
Happy holiday!
Wei
===================================================
Wei Zhang
Department of Biochemistry, Rice Unversity
Email: weizhang@rice.edu
===================================================