Thanks. Good to know it's not just me. :) How might I know which structures do and don't require --use-model? I'm running cif_as_mtz on the whole PDB so I can then run model_vs_data. If I know when to use that switch and when not to, I can mend my script accordingly. Thanks. :) Eric On Wed, Jun 13, 2012 at 10:59 AM, Nathaniel Echols wrote:

On Wed, Jun 13, 2012 at 7:36 AM, Eric Williams wrote:

...

I'm trying to run phenix.cif_as_mtz on a large number of PDB entries, and several entries give me an assertion error. The example below is from 8icz, but the error always seems to be the same.

This is a bug (which one of us will fix ASAP), but if you omit the --use-model argument, it should avoid the step that's crashing. I haven't come across too many entries in the PDB that actually require this, especially among newer structures.

-Nat _______________________________________________ phenixbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/phenixbb

The work-around seems to work...er...around. Thanks. :) Eric On Wed, Jun 13, 2012 at 11:17 AM, Nathaniel Echols wrote:

On Wed, Jun 13, 2012 at 8:05 AM, Eric Williams wrote:

...

How might I know which structures do and don't require --use-model? I'm running cif_as_mtz on the whole PDB so I can then run model_vs_data. If I know when to use that switch and when not to, I can mend my script accordingly.

Pavel might know this, since he's the person who normally runs this. Normally, I would also suggest that you take advantage of a resource in Phenix:

$PHENIX/chem_data/polygon_data/all_mvd.pickle

which contains model_vs_data results for all applicable PDB entries. However, this has not been updated since at least last August, and in the meantime, there have been thousands of new entries added to the PDB, plus Pavel recently changed the bulk solvent correction and scaling procedure, which tends to result in slightly lower R-factors, so I think it's officially obsolete right now.

As it happens, phenix.cif_as_mtz appears to offer an additional way around the crash, which is the argument "--remove_systematic_absences". Try adding this to your existing script (keeping "--use_model") and see if that fixes the problem. Just to be safe, I would also add "--map_to_asu" (which will move reflections around so the h,k,l indices are in the canonical setting) and "--merge" (which will merge non-unique reflections - keeping Friedel pairs separate, of course). Looks like we need to update our documentation.

-Nat _______________________________________________ phenixbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/phenixbb

Hi Eric, data sets in PDB are not super clean, for example: - there is a whole lot of entries labeled as intensities but in fact amplitudes, same true the other way around; - some neutron structures annotated as X-ray; - the worst case (I forgot the PDB id) is the neutron structure called as X-ray with intensities labeled as amplitudes. So, the --use-model option tries to deconvolute these cases simply by trying all possibilities and scoring the results by R-factor. It makes phenix.cif_as_mtz to run a bit longer. If you really know what you are doing then probably not using --use-model is totally ok. Pavel On 6/13/12 8:05 AM, Eric Williams wrote:

Thanks. Good to know it's not just me. :)

How might I know which structures do and don't require --use-model? I'm running cif_as_mtz on the whole PDB so I can then run model_vs_data. If I know when to use that switch and when not to, I can mend my script accordingly.

Thanks. :)

Eric

On Wed, Jun 13, 2012 at 10:59 AM, Nathaniel Echols mailto:[email protected]> wrote:

On Wed, Jun 13, 2012 at 7:36 AM, Eric Williams mailto:[email protected]> wrote: > I'm trying to run phenix.cif_as_mtz on a large number of PDB entries, and > several entries give me an assertion error. The example below is from 8icz, > but the error always seems to be the same.

This is a bug (which one of us will fix ASAP), but if you omit the --use-model argument, it should avoid the step that's crashing. I haven't come across too many entries in the PDB that actually require this, especially among newer structures.

-Nat

On Thu, Jun 14, 2012 at 9:13 AM, Eric Williams wrote:

Might this also be a bug? When processing PDB entry 4a0e, I get "CifParserError: Wrong number of data items for loop containing _symmetry_equiv.id".

Richard's reply got swallowed (wrong email address!), but here is the answer: ---------- Forwarded message ---------- From: Richard Gildea This is an error in the structure factor CIF itself, the problem is the CIF contains an "empty" loop, which is invalid syntax. We have notified the PDB regarding numerous files with similar errors. It doesn't look like the corrected files have been made available yet. If you remove the following lines from the CIF, then you should not see the problem any further: loop_ _symmetry_equiv.id _symmetry_equiv.pos_as_xyz Cheers, Richard