Hi, I recently noticed that the geometry in my structure is much worse than my previous model. Here are some examples. The RMSD results after a refinement by phenix 1.3b. stage angl bond chir dihe plan repu 3_bss: 0.711 0.009 0.059 18.282 0.002 3.917 after switching to cci version 2008_02_08_1610, I tried to do a geometry minimization by setting wxc_scale = 0. The RMSD I got from the previous model is as following. stage angl bond chir dihe plan repu 0 : 1.394 0.124 0.147 20.782 0.004 3.925 and after 6 macro cycles of geometry minimization, stage angl bond chir dihe plan repu 6_bss: 0.622 0.011 0.120 13.143 0.003 3.822 It funny that the RMSD are so different for the same model. What happened? Jianghai +++++++++++++++++++++++++++++++ Jianghai Zhu, Ph.D Immune Disease Institute Department of Pathology Harvard Medical School 200 Longwood Ave., Boston, MA 02115 Ph: 617-278-3211 Fx: 618-278-3232 +++++++++++++++++++++++++++++++
OK, there is more. Apparently the huge difference between the rmsd bond came from phenix.reduce. After adding the riding hydrogens by phenix.reduce, the rmsd bond jumped a lot. Shouldn't the hydrogen atoms be added with perfect geometry? Jianghai On Feb 21, 2008, at 11:24 AM, Jianghai Zhu wrote:
Shouldn't the hydrogen atoms be added with perfect geometry?
Hi Jianghai, phenix.refine uses the monomer library definitions, phenix.reduce has its own internal database. Since nobody really knows what "perfect" is, these (probably) don't agree exactly (but I haven't actually checked). BTW: older versions of phenix.refine reported mean deviations instead of root-mean-squared deviations. This could be another source of mismatching numbers. Ralf
Hi All, the difference is most likely because of H atoms, indeed (deferent "ideal" values used by phenix.reduce and Monomer Library that is used in phenix.refine). In the next version of phenix.refine the geometry of H atoms will always be regularized (unless free refinement is requested at ultra-high resolution), so it will be always consistent with the Monomer Library definitions. Also, I think H atoms should be excluded from geometry statistics calculations if riding model is used. I will add this as well. If all you need is to regularize the geometry, I would use more specific tool for this: % phenix.pdbtools model.pdb --geometry-regularization You can do it with phenix.refine as well (using wxc_scale=0 and strategy=individual_sites), but don't expect the same outcomes since the default values of number of iterations and number of macro-cycles are different. Also, doing it through phenix.refine will be slower. Pavel.
Why not just the same value to start with?
Also, I think H atoms should be excluded from geometry statistics calculations if riding model is used. I will add this as well.
Then you need to change the geometry statistics output by phenix.refine when riding H atoms are present. Jianghai
Because some programs tend to distort hydrogen's geometry completely (for example, COOT after doing local RS fit) so you start refinement with nonsense H positions. And since in riding model X-H distances are fixed you keep having wrong values. Many users reported this problem. Pavel. I
Is it possible that phenix.pdbtools ignored the H atoms when calculating the geometry statistics?
Currently no. But you can easily do it in two steps: 1) phenix.pdbtools model.pdb remove="element H" output.file_name=model_noH.pdb 2) phenix.pdbtools model_noH.pdb --show-geometry-statistics Pavel.
Here are the results. phenix.pdbtools, riding H in the model. |-Geometry statistics---------------------------------------------------------| | Type | Count | Deviation from ideal | Targets | Target (sum) | | | | rmsd max min | | | | bond | 99522 | 0.004 0.085 0.000 | 0.028 | | | angle | 179035 | 0.793 21.106 0.000 | 0.099 | | | chirality | 7686 | 0.126 4.889 0.000 | 0.396 | 0.060 | | planarity | 15814 | 0.004 0.089 0.000 | 0.060 | | | dihedral | 24985 | 15.478 87.363 0.000 | 1.129 | | | nonbonded | 99522 | 3.939 4.900 1.125 | 0.725 | | | -----------------------------------------------------------------------------| phenix.pdbtools, without H in the model. |-Geometry statistics---------------------------------------------------------| | Type | Count | Deviation from ideal | Targets | Target (sum) | | | | rmsd max min | | | | bond | 51167 | 0.005 0.085 0.000 | 0.054 | | | angle | 69163 | 1.030 21.106 0.000 | 0.194 | | | chirality | 7686 | 0.126 4.889 0.000 | 0.396 | 0.142 | | planarity | 9110 | 0.005 0.089 0.000 | 0.104 | | | dihedral | 18756 | 17.864 87.363 0.000 | 1.504 | | | nonbonded | 51167 | 4.092 4.900 1.434 | 0.639 | | | -----------------------------------------------------------------------------| phenix.refine reports at the end of the refinement with riding H. stage angl bond chir dihe plan repu geom_target 0 : 1.394 0.124 0.147 20.782 0.004 3.925 1.1922e+00 ... 6_bss: 0.622 0.011 0.120 13.143 0.003 3.822 2.6112e-02 They are all different. Jianghai On Feb 21, 2008, at 6:01 PM, Pavel Afonine wrote:
Hi Jianghai, I'm confused about what you're comparing with what. Could you please try this: 1) Take a model; 2) Run phenix.refine with it: % phenix.refine model.pdb data.mtz main.number_of_macro_cycles=0 strategy=none 3) Run phenix.pdbtools with it: % phenix.pdbtools model.pdb --show-geometry-statistics Is the geometry statistics reported by phenix.refine and phenix.pdbtools different? If it is different then there is a problem and to fix it I will need to look at that PDB file. Pavel. On 2/21/2008 5:19 PM, Jianghai Zhu wrote:
Sorry, I should have made it clear. I started a refinement using phenix.refine with riding H in my model. After the refinement, I got the following at the end of the log and the header of the pdb file. stage angl bond chir dihe plan repu geom_target 0 : 1.394 0.124 0.147 20.782 0.004 3.925 1.1922e+00 ... 6_bss: 0.622 0.011 0.120 13.143 0.003 3.822 2.6112e-02 Using this refined model, I did phenix.pdbtools model.pdb --show- geometry-statistics. And I got the following. |-Geometry statistics---------------------------------------------------------| | Type | Count | Deviation from ideal | Targets | Target (sum) | | | | rmsd max min | | | | bond | 99522 | 0.004 0.085 0.000 | 0.028 | | | angle | 179035 | 0.793 21.106 0.000 | 0.099 | | | chirality | 7686 | 0.126 4.889 0.000 | 0.396 | 0.060 | | planarity | 15814 | 0.004 0.089 0.000 | 0.060 | | | dihedral | 24985 | 15.478 87.363 0.000 | 1.129 | | | nonbonded | 99522 | 3.939 4.900 1.125 | 0.725 | | | -----------------------------------------------------------------------------| Using this refined model, I did phenix.refine model.pdb data.mtz main.number_of_macro_cycles=0 strategy=none. And I got the following. ------------------------------------------------------------------------ stage angl bond chir dihe plan repu geom_target 0 : 0.608 0.003 0.120 12.395 0.003 3.928 2.9651e-02 ------------------------------------------------------------------------ After removing the H from the refined model, I did phenix.pdbtools model_noH.pdb --show-geometry-statistics. And I got the following. |-Geometry statistics---------------------------------------------------------| | Type | Count | Deviation from ideal | Targets | Target (sum) | | | | rmsd max min | | | | bond | 51167 | 0.005 0.085 0.000 | 0.054 | | | angle | 69163 | 1.030 21.106 0.000 | 0.194 | | | chirality | 7686 | 0.126 4.889 0.000 | 0.396 | 0.142 | | planarity | 9110 | 0.005 0.089 0.000 | 0.104 | | | dihedral | 18756 | 17.864 87.363 0.000 | 1.504 | | | nonbonded | 51167 | 4.092 4.900 1.434 | 0.639 | | | -----------------------------------------------------------------------------| Jianghai On Feb 21, 2008, at 8:35 PM, Pavel Afonine wrote:
Now I see what you did. I tried exact same things myself using an arbitrary structure from PDB. I couldn't reproduce your problem using both: the current development code and the latest CCI Apps: 2008_02_08_1610. In this case I can only work on this problem if you send us the reproducer so we can reproduce your problem. Pavel. On 2/21/2008 6:06 PM, Jianghai Zhu wrote:
participants (4)
-
Jianghai Zhu -
Pavel Afonine -
Pavel Afonine
-
Ralf W. Grosse-Kunstleve