[phenixbb] refinement of twinning dataset

Nathaniel Echols nechols at lbl.gov
Tue Jun 14 15:05:12 PDT 2011

On Tue, Jun 14, 2011 at 2:54 PM, Ms Chiung-Wen Chang
<chiungwen.chang at uqconnect.edu.au> wrote:
> i have some issues with the refinement of my twinning dataset
> the dataset has 2.1 A resolution
> the twin fraction is around 0.43
> i did not detwin and ran the refinement directly (with twinning refinement)
> The attached is the polygon
> so far the R: 0.1905   Rfree: 0.21
> RMSD angles: 0.91   bonds: 0.006
> clashscore: 28.176
> average B: 45.8
> are the clashscore and Average B in the reasonable range ? because the pattern of the polygon is looked wired

Average B is on the high side, but it's within the observed
distribution.  I wouldn't worry about it unless there are other
problems with the structure.  The clashscore is definitely higher than
it should be, but not so high that it will be terribly difficult to
fix.  The polygon looks particularly weird because your bonds and
angles are unusually tight - which probably isn't a bad thing, but you
could get away with a slightly higher X-ray/stereochemistry weight if
you wanted.  (I'd leave it alone, though - based on the statistics
alone, it sounds like you're nearly done.)

> what can I  do more to get them better ?

Load the output in Coot, and make sure the Probe dots are displayed by
clicking "Re-load Probe dots" in the "Protein" sub-tab of the
"Molprobity" tab in either the phenix.refine or validation GUIs.  You
can also run Probe directly from Coot - it's in the "Validation" menu,
and if you launch Coot from Phenix it will have the correct program
paths set.  Once you have the dots displayed, hopefully it will be
more or less obvious what to fix.  (Keep in mind that they are
calculated with explicit hydrogens, so if you're not usually looking
at these when you build, it may not be obvious why atoms are


More information about the phenixbb mailing list