As far as I know map coeffiicient correspond to detwinned data. But
using detwinned data may not be a good idea.
It would be good to see what are R factor. Another thing to consider
is that different program may use different flags for free R and it
may cause some problem.
What are Rfactors, completeness, percentage of freeR reflections?
These are printed by all programs.
If your solution is wrong and you are using twin refinement even if
you do not have twinned crystals your R factor can be as low as 50%
(that is theoretical limit for random Rfactor when one data are from
twinned and another from untwinned crystals). If you have perfect twin
and you are modelling twin (using twin refinement) then your random
Rfactors can be even smaller.
regards
Garib
On 16 Mar 2009, at 11:51, Eleanor Dodson wrote:
> But dont all twinned refinement programs output detwinned terms for
> a map?
> Certainly REFMAC and SHELX do.
>
>
> Eleanor
>
> Clemens Steegborn wrote:
>> Hi Walter,
>>
>> You should definitely detwin data for map calculation if you have a
>> significant twinning fraction (and only for maps; keep using the
>> twinned
>> data set for refinement). We use the CCP4 program detwin. BUT if
>> Shelxl
>> gives bad density, maybe that's simply what you have, a bad density
>> map -
>> because output from Shelx is already detwinned!
>> BTW, we observed that different programs handled different cases
>> differently
>> well; I would suggest ALWAYS to try more than one program, and also
>> to try
>> Phenix ...
>> Best
>> Clemens
>>
>>
>> -----Ursprüngliche Nachricht-----
>> Von: CCP4 bulletin board [mailto:[log in to unmask]] Im Auftrag
>> von
>> Walter Kim
>> Gesendet: Monday, March 16, 2009 7:22 AM
>> An: [log in to unmask]
>> Betreff: [ccp4bb] Twinned data and maps
>>
>> Hi again,
>>
>> Thanks for your insight into refinement tools for twinned data. I
>> have a
>> couple of twinned data sets that are nearly perfectly
>> pseudomerohedrally
>> twinned. I've begun to refine my data in Refmac5 (using the
>> automated twin
>> refinement), CNS (using the twin inputs) and Shelxl; I'm testing
>> out the
>> different refinement programs to evaluate the best strategy for the
>> refinement. However, I would like to start making maps.
>>
>> 1. Refmac5 - outputs an mtz that is model-biased
>> 2. CNS - maps made via model_map_twin.inp are poor
>> 3. Shelxl - the maps generated in coot from the.fcf file are poor
>>
>> Are there better ways to make cleaner maps with my twinnned data
>> that are
>> less model-biased that I can try to build into? Should I detwin the
>> data and
>> make maps from that (but continue to refine against the twinned
>> data)?
>>
>> Thanks,
>> Walter
>>
>>
>>
>>
>
>
|