[phenixbb] Please help with neutron composite SA-omit

Pavel Afonine PAfonine at lbl.gov
Tue Feb 24 08:31:21 PST 2009


Hi Anna,

a "kick map" is computed as following: we create a large ensemble of 
structures (~500 structures) that are all randomly shaken. The shake 
amount (rmsd distortion introduced to coordinates) varies from 0 to 
1.0A. Then for each structure the map is computed (mFo-DFc or 2mFo-DFc 
or any other requested by the user). And finally, all maps are averaged 
out to produce one "averaged kick map".

Obviously, such map is supposed to have less or no bias, less noise, and 
potentially can clear up some initially bad densities.

The is currently no paper published about this method; we are preparing 
the manuscript. There are scattered references and uses of AK maps, 
mostly by Dusan Turk (he originally implemented it in his program MAIN).

Documentation is not yet updated, this is true. Sorry.

In phenix.refine GUI you can just check the corresponding box to compute 
a kick map (under map menu).
In command line version you need to supply a parameter file that will 
define such map:

refinement.electron_density_maps {
    map {
      mtz_label_amplitudes = 2FOFCWT
      mtz_label_phases = PH2FOFCWT
      likelihood_weighted = True
      obs_factor = 2
      calc_factor = 1
      kicked = True
      fill_missing_f_obs_with_weighted_f_model = True
    }
    map {
      mtz_label_amplitudes = FOFCWT
      mtz_label_phases = PHFOFCWT
      likelihood_weighted = True
      obs_factor = 1
      calc_factor = 1
      kicked = True
      fill_missing_f_obs_with_weighted_f_model = True
    }
}

I was recently trying this with neutron maps and the maps are 
significantly better (computed with AK approach).

Please use the latest version of CCI Apps for this.

Let me know if you have any questions or problems.

Pavel.


On 2/24/09 7:58 AM, Anna Gardberg wrote:
> Hi, Pavel.
>
> I'm not trying to bring up features, exactly, but rather trying to get
> an unbiased, quantitative measure of the the deuterium atoms' nuclear
> density. The idea is to compare the visibility of hydrogen or
> deuterium atoms with x-ray and neutron diffraction, correlating this
> parameter (perhaps in the form of OMIT density) to others.
>
> I have only briefly toyed with kick-maps, but the lack of
> documentation (http://phenix-online.org/documentation/refinement.htm
> doesn't say much about how best to use them) made it frustrating. Can
> you recommend an article about kick maps? Or some suggestions on how
> to best use this feature, please?
>
>
> -Anna
>
> On Mon, Feb 23, 2009 at 11:46 PM, Pavel Afonine <PAfonine at lbl.gov> wrote:
>   
>> Hi Anna,
>>
>> you can ask phenix.refine produce "kick maps" that are supposed to be less
>> biased, contain less noise and, if you are lucky, might bring some more
>> features (that were buried in noise before).
>>
>> Let me know if you have any questions.
>>
>> Pavel.
>>
>>
>> On 2/23/09 11:40 AM, Anna Gardberg wrote:
>>     
>>> Hello.
>>>
>>> I am attempting to create an SA-omit composite map with neutron
>>> scatterers.
>>>
>>> Unfortunately,
>>>
>>>  phenix.autobuild data= "neutron.mtz" model="neutron_004.pdb"
>>> composite_omit_type=sa_omit scattering_table=neutron
>>>
>>> results in an error:
>>>
>>> "Sorry, unknown file or keyword: scattering_table=neutron
>>> Possibilities... (use 'phenix.autobuild --help all' to get full help):
>>>  general.resolve_pattern_command_list=neutron
>>>  density_modification.use_resolve_pattern=neutron
>>>  decision_making.semi_acceptable_r=neutron
>>>  decision_making.acceptable_r=neutron
>>>  multiple_models.multiple_models_starting_resolution=neutron
>>> "
>>>
>>> Can someone here help, please?
>>>
>>> (For what it's worth: I'd use nCNS, but while CNS 1.21 (precompiled
>>> binaries) works on my computer, CNS 1.1 version seems to compile, it
>>> will not run. When I apply the nCNS patch to CNS 1.1, again it seems
>>> to compile, but will not run. )
>>>
>>> Regards,
>>> Anna Gardberg
>>> _______________________________________________
>>> phenixbb mailing list
>>> phenixbb at phenix-online.org
>>> http://www.phenix-online.org/mailman/listinfo/phenixbb
>>>
>>>       



More information about the phenixbb mailing list