Hi James,

The chiral validation in Phenix is looking at chiral volumes.  I believe these are calculated using dihedrals.  The chiral volume value is signed, and will change sign if the handedness of the chiral center switches.

One of the basic text outputs of phenix.chiral_validation looks like this:
  atoms                   ideal    model    delta   sigma  residual   deviation
   A 101  THR  CB
   A 101  THR  CA
   A 101  THR  OG1
   A 101  THR  CG2         2.55     1.70     0.85  2.00e-01  1.82e+01   4.3*sigma

This lists the 4 atoms involved in calculating the chiral volume, starting with the chiral center in question (CB here).
Ideal is the target value, found in the .cif's in geostd
Model is the measured value for the chiral center in question
Delta is the difference between ideal and model
Sigma is is esd value associated with the target, also found in the .cif
Deviation is the number of sigmas away from the Ideal that the Model is

If you use the json=True flag, you can get some additional output that attempts to categorize chiral volume outliers, in the "outlier_type" field. This additional output also appears on MolProbity and in the Phenix GUI.

There are 3 categories of chiral volume outlier:
-"Chiral handedness swaps" are true chiral switches like DAL for ALA. These have a large abs(Deviation) (>10*sigma)

-"Tetrahedral geometry outliers" are squashed or flattened chiral centers. Usually the result of geometry distortions or strain, much like bond length or bond angle outliers. I believe these have a abs(Deviation) between 4 and 10. Their geometry is distorted enough that MolProbity is hesitent to make anyclaim about what the intended handedness was.

-"Pseudochiral naming errors" are for atom centers that appear chiral due to chemical naming conventions, but which are not chemically chiral. Iron-sulfur clusters are vulnerable to this sort of mis-naming. As are the P of DNA/RNA backbone (if OP1 and OP2 names are swapped), the CB of VAL (CG1/CG2) and the CG of LEU (CD1/CD2). These represent a bookkeeping error (that should be fixed) rather than a modeling error.

A explanation with KiNG visuals here: http://molprobity.biochem.duke.edu/help/validation_options/validation_options.html#bondgeometry

Hope that helps.  And thank you for noticing that the help text is incorrect.

Regards,
-Christopher Williams
---Richardson Lab, Duke University


On Sun, Nov 3, 2024 at 11:18 AM James Holton <jmholton@lbl.gov> wrote:
Can anyone tell me the units of the numbers being displayed by
phenix.chiral_validation?  The --help documentation doesn't mention
anything about chirality. Seems to be a copy of the documentation for
omegalyze?

Just thought that might be of interest, and I also want to know what the
numbers mean.

Cheers,

-James

_______________________________________________
cctbxbb mailing list -- cctbxbb@phenix-online.org
To unsubscribe send an email to cctbxbb-leave@phenix-online.org