[sharp-discuss] autosharp not reading column headers in mtz files

[Phil Evans]

further comments, not immediately relevant

On 17 Aug 2012, at 10:26, Clemens Vonrhein wrote:

> Hi Phil et al,

> ...

> ------------------------------------------------------------------
> 
> The easiest solution is to not using MTZ files in autoSHARP but rather
> *.sca files. Those can easily be written out by SCALA or AIMLESS in
> addition to the MTZ files. Just add
> 
>  OUTPUT POLISH    # for SCALA
> 
> or
> 
>  OUTPUT SCALEPACK # for AIMLESS (does POLISH also work?)
> 
> The advantage of using *.sca files in autoSHARP:
> 
> * SHELXD can read them directly.
> 
>   So we don't need to convert back from F/SIGF (in MTZ) to I/SIGI (in
>   a temporary *.sca file) with some subtle differences: the F's will
>   have gone through some truncating procedure and just squaring them
>   doesn't give you back the original intensities (which could be
>   negative).
> 

The MTZ files will normally contain I+ & I- as well as F+ & F-. Do you use Is or Fs (which ShelX needs)? If you read Is from .sca file, how do you convert it to F?


> * you save yourself some typing: no need to give column labels in the
>   autoSHARP input form
> 

But you do need to select multiple files for MAD. Also autoSharp _could_ make intelligent guesses for the columns :-)

> Some additional tips that could be relevant to this:
> 
> * make sure you have the correct scalepack format: the 'merged'
>   version is known to work correctly - see
> 
>     http://www.globalphasing.com/sharp/manual/autoSHARP1.html#datafiles
> 
> * make sure you have anomalous data in them (if it is required for
>   SIRAS/MIRAS or SAD/MAD)
> 
> * it is very helpful to give your MAD wavelengths good identifiers:
>   infl, hrem, lrem or peak
> 
>   The reason: SHELXC doesn't use f'/f" values to distinguish the
>   different wavelengths and their corresponding atomic scattering
>   properties. It wants names like "HREM", "LREM", "PEAK" or
>   "INFL". autoSHARP tries to figure these names out from the f'/f"
>   values that you give in the interface - but this can often be
>   wrong. So by giving those identifier names, you are helping
>   autoSHARP to tell SHELXC the right names. See also
> 
>     http://www.globalphasing.com/sharp/manual/autoSHARP2.html#iden
> 
>   Of course, you should still give the correct f'/f" values
>   (especially for wavelengths close to the edge) - from a good
>   fluorescence scan.
> 

The MTZ file has the advantage of including the wavelength, though you still need f'/f" near the edge