phenix.molprobity geo
Hello, I'd like to compare the restraints generated by Coot to the validation provided by molprobity. When I run phenix.molprobity on my model, it gives me (amongst other things) this: ----------------------- Histogram of nonbonded interaction distances: 1.87 - 2.42: 526 2.42 - 2.96: 3217 2.96 - 3.51: 3639 3.51 - 4.05: 5026 4.05 - 4.60: 6847 Nonbonded interactions: 19255 Sorted by model distance: nonbonded pdb=" O LEU A 21 " pdb=" H GLY A 26 " model vdw 1.874 1.850 nonbonded pdb=" H ILE A 71 " pdb=" O TYR A 80 " model vdw 1.920 1.850 nonbonded pdb=" H GLY A 34 " pdb=" O GLU A 54 " model vdw 1.925 1.850 nonbonded pdb=" O ILE A 71 " pdb=" H TYR A 80 " model vdw 1.937 1.850 nonbonded pdb=" HE ARG A 40 " pdb=" HG1 THR A 76 " model vdw sym.op. 1.955 2.100 -x+3/2,-y,z+1/2 ... (remaining 19250 not shown) NOTE: a complete listing of the restraints can be obtained by requesting output of .geo file. -------------------------- I don't understand this lilst. Should I be concerned by those interactions? Which of these atom pairs are contributing to the clashscore? Perhaps the details are in the .geo file? How do I get a .geo file? I googled, I didn't find the answer. How could I have found the answer without asking here? Thanks, Paul.
Hi Paul,
As I understand it, a .geo file is most usually produced as one of the
outputs of refinement. A few different places, such as this documentation (
https://www.phenix-online.org/documentation/file_formats.html#geometry-restr...)
and this newsletter article (
http://phenix-online.org/newsletter/CCN_2016_01.pdf#page=10) mention this
method of generation.
However, you can generate a geo file directly from a model with this
commandline:
phenix.pdb_interpretation *your_model.pdb* write_geo_files=True
This is probably the more useful option for your purposes.
For understanding clashes, I recommend running phenix.clashscore and using
its dedicated and complete output. This will return a list of all the
clashes, sorted in order of severity.
As to non-bonded interaction distances, I believe that they are close
approaches between atoms that are not covalently bonded. As you can see
from the histogram bins (1.87 - 2.42, 2.42 - 2.96, 2.96 - 3.51, 3.51 -
4.05, 4.05 - 4.60A), these distances do not necessarily indicate clashes,
but instead cover a range of distances in the general vdW contact region.
The distribution of this histogram gives a general sense of how well packed
the model is.
Very close non-bonded interactions may register as clashes in Probe and/or
may generate various error messages in Phenix depending on your options.
For various reasons, there is not an exact mapping between close non-bonded
interactions and clashes.
Hopefully someone with a better understanding of the nonbonded distances
will be able to chime in with a better explanation.
Good luck,
-Christopher Williams
---Richardson Lab, Duke University
On Thu, Mar 4, 2021 at 1:00 PM Paul Emsley
Hi Christopher, Thanks for your informative and prompt reply. On reexamining the output, I think then that I misunderstood the output. The NBCs listed below are from *refinement* analysis, not molprobity analysis. I was mislead because there were no NBCs listed in the Molprobity part of the log file. Yes, you are right, the output of phenix.clashscore was what I actually wanted, not phenix.molprobity and a .geo file. Regards, Paul. On 04/03/2021 18:51, Christopher Williams wrote:
participants (2)
-
Christopher Williams -
Paul Emsley