[phenixbb] How to extend R-free set to higher resolution by new phenix-1.5.2

[Thomas C. Terwilliger]

Hi PC,

>>
>> Hi,
>>
>> I know this topic was discussed before, but I am still confused and still
>> unable to refine my map merged from hires data because phenix.refine
>> complains "No array of R-free flags found"
>>
>> The command was used to generate a hires map.
>> phenix.autobuild data=resolve_1.mtz hires_file=hires.sca
>> \seq_file=sequence.dat maps_only=True
>>
>> So, now I have "exptl_fobs_phases_freeR_flags.mtz", "cycle_best_1.mtz",
>> "overall_best_denmod_map_coeffs.mtz"...several files in my autobuild
>> folder,

What this command has done is to take the phases in resolve_1.mtz and
combined them with amplitudes from hires.sca and added free R flags to
generate exptl_fobs_phases_freeR_flags.mtz .

Additionally the file overall_best_denmod_map_coeffs.mtz now contains the
above amplitudes along with some kind of density-modified phases.

I would caution that this is probably not what you really wanted to do. 
Instead:

Almost always you should use for "data=" a file that contains experimental
phases only (not density-modified phases).  So it would be better to use
the command:

phenix.autobuild data=phaser_1.mtz hires_file=hires.sca \
   eq_file=sequence.dat maps_only=True

which would instead density modify using hl coeffs from modify
phaser_1.mtz, extending phase information to the data in hires.sca .

Now you will have exptl_fobs_phases_freeR_flags.mtz with
(a) experimental amplitudes from hires.sca
(b) experimental phases and hl coefficients (where available) from
phaser_1.mtz
(c) free R flags to the full resolution of hires.sca

and overall_best_denmod_map_coeffs.mtz  with coefficients for a
density-modified map at the resolution of hires.sca.

>>
>> How do I merge low and high resolution data by
>> "phenix.reflection_file_editor"?  What files (input and output arrays)
>> should I choose?

I think reflection_file_editor will only copy arrays (or columns) of data,
and not merge several columns of data into one.  Note that merging high
and low resolution data requires (1) a scale factor between the datasets
and (2) possibly a b-factor or anisotropic b-factor relating the two
datasets. It cannot be properly done simply by merging reflection files.

The reflection_flle_editor will add free R flags if they are missing or
extend them for you if they are present only to low resolution, however. 
This allows you to take a low-res dataset with free R flags and a high-res
dataset with no free R and to transfer the existing low-res free R flags
to the high-res dataset and to extend them to full resolution.

If you want, you can use phenix.solve to scale low and high resolution
data by pretending that they are part of a mad dataset:

make a file "scale.csh" like this, after editing to match your case:

#!/bin/csh -f
cp $PHENIX/solve_resolve/ext_ref_files/c2.sym .
phenix.solve<<EOD
resolution 20 3.0
cell 76 28 42 90 103 90
symfile c2.sym
readdenzo    ! readformatted/readdenzo/readtrek/readccp4_unmerged
premerged            ! premerged/ unmerged
read_intensities     ! read_intensities/read_amplitudes
lambda 1
mad_atom c
fprprv_mad 1.
fprimv_mad 1.
rawmadfile low.sca
rawmadfile high.sca
skip_solve
sad
EOD

Then run with
chmod +x scale.csh
./scale.csh

Now solve_1.sca should have your scaled, merged data.


I hope that helps!
-Tom T