[phenixbb] Rfree doesn't decrease and keeps around 36%, High ramachandran outlier, High RMSD

[chen c]

Dear all,

I am solving a structure which diffracted to 2.8 angstrom resolution and
containes 8 molecules (MW:12 kDa) per ASU. It's noteworthy that though the
model used for molecular replacement shares 50% sequence identity with my
protein, however, severe clashes were found as listed below:


   Packing Table
   -------------
   Solutions accepted if total number of clashes <= 5% of trace atoms
      i.e. total number of clashes <= 32
   AND if number of clashes <= 5% of trace atoms for each ensemble
      i.e. ensemble1: number of clashes <= 4
   #    #  #Clash Packs SpaceGroup  Annotation
   1         17    NO   C 2 2 21    ... PAK=11 LLG=802 RFZ=3.8 TFZ=9.3


Only after I changed the packing criteria from 5% to 100%, can I get the
solution.

Severe clashes and even coincidence of different molecules were found as
briefly listed below:

Molecule A (C-terminal)  vs Molecule D (N-terminal) of neighbouring ASU
Molecule B (C-terminal)  vs Molecule C (N-terminal)
Molecule C (C-terminal)  vs Molecule D (C-terminal)
Molecule E (C-terminal)  vs Molecule G (N-terminal)
Molecule F (C-terminal)  vs Molecule H (N-terminal)
Molecule G (C-terminal)  vs Molecule G (C-terminal) of neighbouring ASU
Molecule H (C-terminal)  vs Molecule H (C-terminal) of neighbouring ASU

Upon this, I manually rebuild the structure in COOT and particularly
deleted those coincided atoms.
Then I refined the structure in phenix. However, all the refine gives a bad
result, namely, RMSD(angle) at 1.8-2.1, RMSD(bond) at 0.015-0.02, The
Rwork/Rfree at around 0.28/0.36, Ramachandran outlier + Rotamer outlier at
around 15-20%. I tried a lot of refinement strategy and the current status
is :

RMSD(angle)= 1.545, RMSD(bond) = 0.01, Rwork/Rfree = 0.290.36, Ramachandran
outlier=11.3%,Rotamer outlier =1.5%.

The data set belonged to C2221 and phenix.xtriage analysis reflects no
twinning or pseudotranslation problem. In addition, there are two
beta-turns in each molecule which displays poor electron density and
geometry. And this two loop are the very residues which appears in
disallowed regions in ramachandran diagram apart some other residues.
However, these residues should have little effect on the R factor
considering of their low contribution to the overall structure due to their
poor electron density.

Has anyone has experiences about this? Your help and advice is greatly
appreciated!

Thank you!
Chen


-- 
Cheng Chen, Ph.D. Candidate
Laboratory of Structural Biology
Life Science Building,Tsinghua University
Beijing 100084
China
Tel:+86-10-62772291
Fax:+86-10-62773145
E-mail:chench at xtal.tsinghua.edu.cn

北京市海淀区清华大学生命科学馆201-212室
邮编：100084
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://phenix-online.org/pipermail/phenixbb/attachments/20160512/44aed3e0/attachment.htm>