[phenixbb] phenix.refine Ile peptide bond breaks

SIPPEL,KATHERINE H ksippel at ufl.edu
Wed May 6 18:07:57 PDT 2009

Hi all,

I've got a 2.1 angstrom structure that I am refining.  The space 
group is P212121.  The structure is a dimer which I am refining 
without NCS restraints.  Each monomer has 333 residues and one 
ligand, there are 33 Ile residues in each monomer.  The first 
couple of rounds of refinement I had no problems, but during the 
third round phenix started breaking the main chain at the peptide 
bond and the side chain between CA and CB on some of the Ile 
residues.  Specifically there are two breaks in chain A and four 
breaks in chain B, with two additional residues in chain B 
breaking only the CA-CB bond.  Only one of the breaks is between 
the same residue in both chains.

I'll be boring and run through my refinement process.  This is the 
first time I've used phenix for anything but AutoSol and it is 
very likely that I have messed something up.  Feel free to correct 
anything else I've done wrong along the way.

Round 1, I rigid body refined chain A and B independently, 
performed simulated annealing, and refined individual coordinates. 
 In COOT I manually refined and removed some bad loops.  A few of 
the Ile were deleted but none of the problem ones

Round 2, I refined individual_sites and individual_adp and turned 
off simulated annealing.  In COOT I rebuilt some of the loops and 
fitted the ligand into the density. I generated a .cif file in 

Round 3, I refined the same as round 2 except that I included the 
.cif file.  This is when I got the peptide bond breaks for Ile 
residues.  I figured it was a fluke, fixed the breaks in COOT and 
rebuilt a few more residues in the loops.

I ran another round of refinement and the same breaks showed up in 
the pdb file.  I double checked the .geo file to see if Real Space 
Refine had made the bond distances were too long, but all the 
input distances were within one or two hundredths of an angstrom 
from ideal.

I considered it was maybe the ADP refinement as I was borderline 
pushing the number of parameters I was refining given the number 
of reflections I had.  I turned off the individual_adp refinement 
and reran refine_003.  No luck.

I tried increasing the weight of the geometry restraints over the 
observed data.  I did this by increasing target_weights.wxc_scale 
= 1.5.  (I suspect that I got this wrong.  As I understood the 
wxc_scale is the ratio of the geometry restraints to the observed 
data.  Feel free to point and laugh as long as you also provide an 
explanation as to what this parameter really means and how I 
should have done it,please.)  Still no luck.

For future information none of the problem isoleucines were 
located near the ligand or near any of the loops I was 

I think that's everything.  Any help would be appreciated,



Ph. D. candidate
McKenna Lab
Department of Biochemistry and Molecular Biology
College of Medicine
University of Florida

More information about the phenixbb mailing list