[phenixbb] Problem in reducing R-factor

Pavel Afonine pafonine at lbl.gov
Fri Oct 11 08:27:17 PDT 2013


>   I am solving a crystal structure at 1.3Å.

I would expect Rwork~0.15-0.17, Rfree~0.17-0.19 at this resolution 
(numbers obtained using phenix.r_factor_statistics).

> Completeness of the data-set (...) at this resolution are 99.5%

Completeness reported as single number (as opposed to completeness 
reported vs resolution) isn't really very informative..

> Structure has two molecules per asymmetric unit. I used TLS parameters and NCS restraints.

Resolution limit for using NCS is roughly about 2A. So turning OFF NCS 
is clearly a good idea. However, effect of using torsion NCS should 
vanish at this resolution. So I am particularly interested to know if 
turning NCS off helps or shows any effect at all.

> Rwork and Rfree are 18.09 and 19.62 respectively.

If it is really 1.3A resolution then R-factors could be better.

> I tried with reducing WXC_scale

If you choose to do it manually then at this resolution you should 
rather try increasing this value, as the target is

T = wxc_scale * wxc * Tdata + wc * Trestraints .

If none of the above helps please send me data and model files and I 
will try suggesting optimal refinement strategy.


More information about the phenixbb mailing list