"As a general rule when a deposition comes in where there has been additional processing of the structure factors we ask for either: the raw images to be deposited and the DOI provided to us or the structure factors prior to the additional processing to be added as a second datablock in the MTZ/mmCIF file.  This is to ensure that users have access to the data prior to the additional processing.  Currently STARANISO falls into this additional processing category, partly as it is new, but as techniques become used more widely we do review this and change our practice."

1. If intensities were used as input to the final structure refinement, then mmCIF datablock #1 should contain the exact set of intensities and standard deviations used in the refinement, after whatever diffraction cut-off was applied.  Alternatively, if structure amplitudes were used in the final refinement, then mmCIF datablock #1 should contain the exact amplitudes and standard deviations used as input to that refinement, whether STARANISO or other software (e.g. XDSCONV or CTRUNCATE) was used in the I-to-F conversion step.  The reason for including these data is so that the validation step in the deposition process can test whether the final results can be reproduced from that input.  It will also allow re-refinement using packages such as PDB_REDO.

2. The mmCIF datablock #2 should contain the scaled and merged intensities to which no restrictive diffraction cut-off has been applied, i.e. the data should extend out to the edges of the detector.  Example: the scaled and merged intensity output (e.g. from AIMLESS) when no restrictive diffraction cut-off had been applied to the data at this stage.  This is exemplified by the 'aimless_alldata.mtz' file as produced by autoPROC in the 'unmerged data' protocol of the STARANISO server.

   This is in fact the required input to the STARANISO server when using the 'merged data' protocol.  Note particularly that it is a requirement of the server, whether one submits unmerged data (already scaled or unscaled) or already merged data, that no diffraction cut-off of any kind has been applied, and we think this should also pertain to PDB deposition, since such a cut-off surely falls into the category of "additional processing", and therefore requires that the data prior to such processing be deposited.

3. We recommend that the mmCIF file containing the map coefficients (possibly more than one set) output by the final refinement should be prepended by the user to the mmCIF file output by the server (again using e.g. Unix 'cat'), so that structure-factor calculation will use this dataset, rather than the one from the processing (this is important if there has been some re-indexing between the two datasets).  We think it is important that subsequent users of the deposited data should be able to reproduce exactly any published electron-density maps, in order to verify the authors' claims against their very own maps: this requires that the map coefficients used to produce those maps be deposited.

   NOTE: some refinement programs may mirror the input intensities/amplitudes in the output, so in those cases it is not strictly necessary to include mmCIF datablock #1 above, and it is sufficient to take both the intensities/amplitudes and the map coefficients from their output file.  It obviously does no harm to duplicate the information (provided of course the data are identical), but beware that some programs (e.g. currently REFMAC, though not BUSTER) write a modified version of the input data into their output file; in such a case it would be advisable to include all three (or more) datablocks.  For information on deposition of BUSTER refinement outputs consult the Wiki.

1. If unmerged data were submitted to the STARANISO server, the relevant merging statistics required for deposition are shown in the green box under 'Summary of merging statistics for observed data'.

2. If merged data were submitted, these statistics are not available from the server.

3. Furthermore, the merging statistics obtained for the merged data submitted to the server (e.g. from the AIMLESS logfile) do not adequately describe the data as output by the server in the merged data protocol.

4. For these reasons, it is recommended to submit unmerged data wherever possible to the server (or use autoPROC for processing from the images when e.g. a very large number of datasets are involved).