Re: Amore...(long winded paraphrasing of the manual ;-)

[Lisa Edberg <edberg@onyx.cmc.uab.edu>]

***  For details on how to be removed from this list visit the  ***
***    CCP4 home page http://www.dl.ac.uk/CCP/CCP4/main.html    ***

Dear Balaij,

First of all, one should try the defaults and suggestions in the manual
:-).
The manual is really useful if you set your mind to understanding it.
I know that seems obvious, but usually the defaults work.

Here are some general pointers, obvious, to be sure...
1. Anything ending in "0" is your data as in hklpck0
2. Anything ending in "1" belongs to model "1"
3. Name your output files so you can keep track of them.

There are several steps involved:
1. SORTING: puts your reflection data into amore format to speed up the
calculations.
for your input hklin, it doesn't hurt (or help) to have uniqueified
data. You'll need those free-r flags later anyway. Use all of the data,
no matter what your model has.

[IN: data.mtz  OUT: data_pch.hkl]

2. TABLING: places your model in the optimum position and calculates the
fourier coefficients. Make sure your model.pdb file header includes
CRYST and SCALE cards in the header that MATCH YOUR REFLECTION DATA.
Best to use reasonable B factors (15-20).
Use the lower of model/reflection file high resolution limits (I have
used 2.5A). Most likely you will have to increase TABLING_MI to 1 000
000 or more if you have high resolution data and model. Just keep
increasing it until it spits out a table (model.tab)

[IN model.pdb OUT model_rot_tran.pdb model.tab]

3. ROTING: rotation function
Although the sample input does all of the steps at once, you don't need
to repeat all steps every time.

3A. GENErate step: do this only once- comment it out of the subsequent
runs.
Resolution : use the resolution limits of the TABLING step.
Cell_model: Look in the tabling log file... at the end is the maximal
distance from the center of mass.
I use 70 70 70 for a model with maximal dist 24.90 and minimal box
45.56    40.46    33.43

[IN model.tab OUT model.hkl]

3B : CLMN step calculates the spherical harmonics. CRYST is your
reflection data, MODEL 1 is for your first (only) model. Use various
resolution ranges: 8-4, 7-3.5, 6-3 etc. in your trials. Use SPHERE
~75-80% of your maximal distance. You may want to try various values
here too. I used 19 (75%)

[IN data_pch.hkl OUT data.clmn] and
[IN model.hkl OUT model.clmn]

3C: ROTAte CROSS (crystal vs model) or SELF (crystal vs crystal) for
self, you don't need to generate the model CLMN's or refer to MODEL.
STEP size 1.5 to 2.5 is good- Start with the larger value.
BMAX 90 for orthorhombic. PKLIM should be 0.2, NPIC 50.

the resolution limits and sphere's must match in the CLMN and ROT steps.

Useable CC's will range from 10 to 20% (or 30)

Take the SOLUTIONRC's out of the log files. The same peaks should keep
coming up in the resolution ranges. If you have multiple peaks in the AU
related by the self rotation, the output peaks should be related. If
your multiple peaks are not related by simple NCS, you will have
trouble. Make derivatives. Trust me. Save a big dent in your head.

[IN data.clmn model.clmn OUT cross.map, list of SOLUTIONRC in log]

4. TRAING: start with 1 molecule and search the top 30-50 SOLUTIONRC
peaks. When you find the first solution (grep SOLUTIONTF traing1.log,
sort by CC) add FIX to that line and paste it into the top of the
SOLUTIONRC list, change NMOL to 2 and run again to find the second, etc.
If you have used BMAX 90 in your rotation function, and your second
molecule is related by a twofold close to parallel to the cell axis,
then you must repeat the search on the first rotation solution to find
the second. Use a relatively high resolution limit.

Top CC's will increase to 25-40%. If you search with a correct model
(solved identical structure) CC can be 60-85%, Rfac 40%. The CC will
increase with the second molecule.

 If your third molecule is hard to find, refine the positions of the
first two using FITING before the translation search for the third.

OR MAKE DERIVATIVES AND SAVE THE TROUBLE.

[IN data_pch.hkl, model.tab, list of SOLUTIONRC's,  OUT trans.map, list
of SOLUTIONTF]

5. FITING: to refine the solutions. Use the entire resolution range in
your table. If your solution(s) is (are) clear, then just refine that
one. Else, refine a list of your (15-20)  best solutions (sorted by CC)
and look to see which ones improve drastically.

[IN model.tab, data_pch.hkl, list of SOLUTIONTF; OUT  list of
SOLUTIONF's]

6. Apply the solutions to your model using PDBSET. Use the rotated and
translated output .pdb from TABLING. Transform to the proper AU. Check
to see that they and their symmetry mates pack properly with no serious
overlaps.

[IN model_rot_tran.pdb OUT solution_1.pdb]

I hope this helps.
Lisa

--
Lisa Edberg (205)934-1611 FAX:(205)934-0480
Center for Macromol. Cryst., Univ. Alabama-Birmingham
http://www.cmc.uab.edu/misc/edberg/index.html
We must trust the dark to show us the stars.