Bond Length Outliers (correction)

From: Greg Costakes

Date: 16 February 2012 16:31

I am currently in the final steps of refining a 1.3A structure and am coming across a slight problem. According the the pdb file, I have an Rmsd bond length of 2.55. MolProbity identifies three outliers which correspond to the bond lengths of:

Asp: C--O   , bond length = 1.2A
Arg: C--O   , bond length = 1.15A

Ala: N--Ca  , bond length = 1.43A

Real space refinement in Coot does not help and if I Regularize the zone it completely distorts the backbone. So my question is, how do I fix these bond length outliers? Do I need to be concerned with them? Any advice will be much appreciated. Thank you!

-------------------------------------------------------------------------------

Greg Costakes

PhD Candidate

From: Greg Costakes
Date: 16 February 2012 17:00

Ahh yes, I looked at the wrong line. My Rmsd bond angle is 2.55 degrees (not bond length). MolProbity states that my only abnormal angle is 124.23 degrees between O--C--N of an Arg. Real Space Refinement does not change anything and Regularizing the zone completely distorts the backbone. Any suggestions on how to fix this?

-------------------------------------------------------------------------------
Greg Costakes

Greg,

Your RMSD on bond lengths should be around 0.01A (your structure vs.
idealized library), and the RMSD on bond angles should be around 1.5deg.
You must be using an incorrect value of the weight factor between
structure factors and geometric factors, and relying too heavily on
structure factors.

Bernie



On Thu, February 16, 2012 10:31 am, Greg Costakes wrote:
> I am currently in the final steps of refining a 1.3A structure and am
> coming across a slight problem. According the the pdb file, I have an Rmsd
> bond length of 2.55. MolProbity identifies three outliers which correspond
> to the bond lengths of:
> Asp: C--O , bond length = 1.2A
> Arg: C--O , bond length = 1.15A
> Ala: N--Ca , bond length = 1.43A
>
>
> Real space refinement in Coot does not help and if I Regularize the zone
> it completely distorts the backbone. So my question is, how do I fix these
> bond length outliers? Do I need to be concerned with them? Any advice will
> be much appreciated. Thank you!
>
>
>
> -------------------------------------------------------------------------------
> Greg Costakes
>
>
>

----------
From: Christian Roth
Hi,
If I remember correctly this angle is for the planarity of the peptide bond.
Maybe you have a real deviation which might occur not so seldom than expected.
You have a quite high resolution. If the density is really convincing then you
may accept this outlier. Not every outlier is a mistake. However with 2.55°
the r.m.s.d. for your angles is quite high I think.

Christian


Am Donnerstag 16 Februar 2012 18:00:10 schrieb Greg Costakes:
> ---- Greg Costakes

> >------ Greg Costakes

----------
From: Dale Tronrud
  Using the Protein Geometry Database (pgd.science.oregonstate.edu) I looked
up all Arg residues in models with resolution of 1.3 A or better and found
5920 examples.  The mean value of the O-C-N angle (and I'm assuming that the
O and C atoms are in the Arg) is 122.6 deg with a sigma of 1.1 deg.  338 of
them have a value greater than 124.23 deg, or about 6%.  It doesn't look to
me that this piece of structure is an outlier.

  Regularizing may move atoms out of density but it shouldn't "distort"
anything, it should make it, <cough>, more regular.  If regularizing is
doing something bad there is a problem with the regularizer not the structure.

  Does you model have any ligands that might have horrible angles but not
be reported by MolProbity?

Dale Tronrud

----------
From: Bernhard Rupp (Hofkristallrat a.D.)
Btw, re other sources of deviation: Molprobity does not report geometry deviations beyond CB.  The RUN500 command from CCP4i does.
BR