[phenixbb] PHENIX refine: detached sidechain atoms - SOLUTION

Schubert, Carsten [PRDUS] CSCHUBER at its.jnj.com
Fri Jun 25 05:16:46 PDT 2010


Good to know, that is a nasty side-effect of a complex procedure. We had
a similar case here, which was if I remember correctly was also phenix
version dependent, and we could not make sense of it. The ligand in our
case was actually generated in elbow, so that would not shield one from
problems. It always helps to give the cif files a second look. 


> -----Original Message-----
> From: phenixbb-bounces at phenix-online.org [mailto:phenixbb-
> bounces at phenix-online.org] On Behalf Of Pavel Afonine
> Sent: Friday, June 25, 2010 1:22 AM
> To: oliver.king at chem.ox.ac.uk; PHENIX user mailing list
> Subject: Re: [phenixbb] PHENIX refine: detached sidechain atoms -
> SOLUTION
> 
> Hi Oliver,
> 
> thanks for sending the inputs so we could reproduce this issue and
> figure out what is wrong.
> 
> The problem maker is the CIF file for your ligand that specifies
> unrealistic (too small) esds for some torsional angles that in turn
> creates huge terms in geometry restraints target that finally resulted
> in a huge X-ray weight. Using huge X-ray weight obviously means that
> you
> are doing nearly unrestrained refinement and therefore no surprises
> that
> the geometry was systematically distorted. This is something we will
be
> catching automatically in future, but meanwhile it is a good idea to
> use
> correct ligand CIF files.
> 
> The solution: create a CIF file using either phenix.elbow or, better,
> using phenix.ready_set commands. For example,
> 
> phenix.ready_set model.pdb
> 
> will create a CIF file for all ligands in your model.pdb file. Using
> this file in refinement eliminates the problem that you have reported
> this morning. Also, this command adds H atoms to model.pdb file, that
> you can then use in subsequent refinements.
> 
> I went ahead some more and did some refinement using the strategy that
> I
> think is the best for your case.
> 
> Your starting R-factors (corresponding to the model you sent me):
> r_work = 0.1853 r_free = 0.2523 bonds = 0.069 angles = 4.460
> 
> And after some refinement that I've done:
> r_work = 0.1779 r_free = 0.2202 bonds = 0.015 angles = 1.600
> 
> In refinement I used:
> - automatic water update;
> - TLS;
> - riding hydrogen atoms.
> 
> You can further improve your model by:
> - running automatic rotamer fixing (fix_rotamers=true);
> - weights optimization (optimize_wxc=true optimize_wxu=true);
> - more thoughtful selection of TLS groups.
> 
> I'm sending to you all the files OFFLIST (in the next email).
> 
> Please let me know if you have any questions.
> 
> Pavel.
> 
> 
> 
> 
> On 6/24/10 3:42 AM, Oliver King wrote:
> > Hi All,
> >
> > I've noticed that when refining a model in phenix.refine using a PDB
> > file from Refmac, the sidechain atoms of certain residues become
> > detached from the rest of the molecule and appear to float on their
> > own, at least when viewing in Coot. I think this is down to the
> format
> > of the PDB file. For example the atoms of a Leu residue from a
> > phenix.refine file which displays well is of the form:
> >
> > ATOM      1  N   LEU A   8      50.022 -34.247  -5.817  1.00
> > 58.12           N
> > ATOM      2  CA  LEU A   8      49.788 -34.905  -4.539  1.00
> > 55.84           C
> > ATOM      3  C   LEU A   8      48.339 -35.358  -4.348  1.00
> > 60.39           C
> > ATOM      4  O   LEU A   8      48.008 -36.016  -3.360  1.00
> > 68.40           O
> > ATOM      5  CB  LEU A   8      50.219 -34.011  -3.373  1.00
> > 57.98           C
> > ATOM      6  CG  LEU A   8      51.690 -34.126  -2.975  1.00
> > 65.66           C
> > ATOM      7  CD1 LEU A   8      52.014 -33.170  -1.836  1.00
> > 59.99           C
> > ATOM      8  CD2 LEU A   8      52.020 -35.566  -2.594  1.00
> > 68.57           C
> >
> > where as from a Refmac PDB which also behaves well, it is of the
> form:
> >
> > ATOM      1  N   LEU A   8      50.453 -35.722  -5.617  1.00
> > 20.00           N
> > ATOM      2  CA  LEU A   8      49.649 -35.131  -4.482  1.00
> > 20.00           C
> > ATOM      3  CB  LEU A   8      50.190 -33.735  -4.147  1.00
> > 20.00           C
> > ATOM      4  CG  LEU A   8      51.461 -33.755  -3.275  1.00
> > 20.00           C
> > ATOM      5  CD1 LEU A   8      52.556 -32.768  -3.741  1.00
> > 20.00           C
> > ATOM      6  CD2 LEU A   8      51.082 -33.546  -1.799  1.00
> > 20.00           C
> > ATOM      7  C   LEU A   8      48.166 -35.063  -4.824  1.00
> > 20.00           C
> > ATOM      8  O   LEU A   8      47.822 -34.257  -5.568  1.00
> > 20.00           O
> >
> > but after refinement in phenix it becomes
> >
> > ATOM      1  N   LEU A   8      50.734 -35.936  -5.935  1.00
> > 54.45           N
> > ATOM      2  CA  LEU A   8      49.846 -35.185  -4.892  1.00
> > 51.89           C
> > ATOM      3  CB  LEU A   8      50.377 -33.761  -4.516  1.00
> > 51.22           C
> > ATOM      4  CG  LEU A   8      51.550 -33.730  -3.431  1.00
> > 59.27           C
> > ATOM      5  CD1 LEU A   8      52.457 -32.407  -3.635  1.00
> > 60.88           C
> > ATOM      6  CD2 LEU A   8      51.120 -34.090  -1.879  1.00
> > 55.50           C
> > ATOM      7  C   LEU A   8      48.268 -35.042  -5.221  1.00
> > 47.92           C
> > ATOM      8  O   LEU A   8      47.989 -34.282  -6.195  1.00
> > 47.32           O
> >
> > and the CD2 atom is now too far away to be part of the residue in
> Coot
> >
> > Is there an easy way to convert a PDB from Refmac into one which
will
> > behave itself when put through phenix.refine? I've tried using
> > phenix.pdb_tools and also putting the model through Molprobity and
> > hoping that that the output would be corrected.
> >
> > Thanks,
> >
> > Olly King
> >
> > _______________________________________________
> > phenixbb mailing list
> > phenixbb at phenix-online.org
> > http://phenix-online.org/mailman/listinfo/phenixbb
> _______________________________________________
> phenixbb mailing list
> phenixbb at phenix-online.org
> http://phenix-online.org/mailman/listinfo/phenixbb




More information about the phenixbb mailing list