Re: [ccp4bb]: accuracy and precision of unit cell axes - pdb_extract

[David W Borhani]

Very interesting. I just checked ** mosflm ** postrefinement logs of a few crystals, good and bad. ESD of about 1 in 10,000 for a very good crystal, more typically 1 in 5,000. About 1 in 1,000 for a bad one (e.g. 137.71 +/- 0.12 A).

Surely someone in the distant past has compared post-refinement of a "good"protein crystal (hundreds if not thousands of reflections contributing to the refinement) vs. the typical 20 or 25 high-angle reflections in a small molecule crystal cell refinement in terms of accuracy & precision of the results? (I should know this, but, is a small mol. cell refinement independent of the detector distance, as is postrefinement?)

Dave



Hi.

I've found the discussion of significant digits and esds for
the unit cell constants very interesting, but the numbers you
listed from HKL2000 really got my attention.  The last time
I had any faith in the esds for unit cell constants was 30 years
ago when working with stable, reproducible small molecule crystals.
We could center on high angle reflections, refine the unit cell
constants, and get esds with a precision of a part in 10,000.
When we did that, we were pretty proud of the precision, but I'm
not sure many believed the numbers were that accurate.

The values you report in your example are a part in 180,000,
and that's stupendous!  Can you, or any of the other folks who
understand where the esds come from in HKL2000, explain why those
measurements are so much more precise now?

Thanks.

Ron


On Mon, 5 Jun 2006, Dunten, Pete W. wrote:

> ***  For details on how to be removed from this list visit the  ***
> ***          CCP4 home page http://www.ccp4.ac.uk         ***
>
>
>
> Actually, if you use pdb_extract to prepare your deposition,
> and for example have used HKL2000 for the integration and scaling,
> you will get something like this, with esd's included -
>
> _cell.length_a          87.453
> _cell.length_b          146.925
> _cell.length_c          189.787
> _cell.angle_alpha       90.000
> _cell.angle_beta        90.000
> _cell.angle_gamma       90.000
> _cell.length_a_esd      0.001
> _cell.length_b_esd      0.002
> _cell.length_c_esd      0.001
> _cell.angle_alpha_esd   0.0
> _cell.angle_beta_esd    0.0
> _cell.angle_gamma_esd   0.0
> _cell.volume            2438570.0
>
> Wake up to the new world of mmcif!  You can capture a lot more
> information in your depostions now.  See http://pdb-extract.rcsb.org/
>
> All the best, Pete
>
>> David Borhani wrote
>>>
>>> Marilyn,If you have post-refined the data (likely; default in HKL2000 & mosflm), then you should have
>>> standard deviations reported for the cell axes, e.g. a = 51.145 +/- 0.125 Angstrom. Since the
>>> PDB doesn't have a way (that I know of) to report the SD . . .
>
>
>
>
>
>