Scattering from riding hydrogens?
Hello, I have a question regarding two parameters and how they affect riding hydrogens. contribute_to_f_calc = True high_resolution_limit_to_include_scattering_from_h = 1.6 At lower resolution, setting contribute_to_f_calc = false is frequently helpful when using riding hydrogens and has been discussed before. But what is high_resolution_limit_to_include_scattering_from_h = 1.6 doing? If scattering from hydrogens is only included at higher than 1.6A resolution, it would seem like contribute_to_f_calc = True/False would not have any effect at say, 3.5A, but comparing identical refinement runs with contribute_to_f_calc = True or contribute_to_f_calc = False clearly gives different results. Can hydrogens contribute to Fcalc in a way aside from their scattering, or is their contribution accounted for independent of the position of the riding hydrogen (e.g., additional scattering term added to the associated heavy atom)? I think I am misunderstanding what these two parameters are doing and would appreciate some clarification. Thanks, Damian Ekiert
Hi Damian, the contribution of H atoms to the total model structure factor Fmodel is refined in recent PHENIX versions, so these two parameters are not necessary and in fact are leftovers that I forgot to remove when switching to the new code. I will do it for the newer versions. Sorry for the confusion. So, if you have H atoms in your input PDB file then you don't have to change any parameter. If you don't want to use H atoms then simply make sure they are not present in input PDB file. Current implementation of handling of hydrogen atoms in refinement is partially inspired by J. Appl. Cryst. (2010). 43, 1100-1107 and thoroughly tested by various numerical tests (that we should probably publish in one of CCN issues). Pavel On 11/14/11 2:21 PM, Damian Ekiert wrote:
Hello,
I have a question regarding two parameters and how they affect riding hydrogens.
contribute_to_f_calc = True high_resolution_limit_to_include_scattering_from_h = 1.6
At lower resolution, setting contribute_to_f_calc = false is frequently helpful when using riding hydrogens and has been discussed before. But what is high_resolution_limit_to_include_scattering_from_h = 1.6 doing? If scattering from hydrogens is only included at higher than 1.6A resolution, it would seem like contribute_to_f_calc = True/False would not have any effect at say, 3.5A, but comparing identical refinement runs with contribute_to_f_calc = True or contribute_to_f_calc = False clearly gives different results. Can hydrogens contribute to Fcalc in a way aside from their scattering, or is their contribution accounted for independent of the position of the riding hydrogen (e.g., additional scattering term added to the associated heavy atom)? I think I am misunderstanding what these two parameters are doing and would appreciate some clarification.
Thanks,
Damian Ekiert _______________________________________________ phenixbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/phenixbb
Pavel, Thanks for your response. This question was motivated by discussions of whether or not to deposit riding hydrogens in the PDB, so perhaps I can ask a more general question. As I understand things, the positions of the riding hydrogens (from an input PDB file) would would be incorporated into the calculation of Fcalc (I guess automatically now and before when contribute_to_f_calc = True). If this is correct, then removing riding hydrogens before deposition seems very problematic, as they were effectively used in the calculation of R/R-free, etc. However, if the contribution of hydrogens is more abstract (e.g., the riding hydrogens are not used in the calculation, but the presence of hydrogens in the protein in the crystal are accounted for in some other way), leaving them out of the deposition is somewhat less of a problem (setting aside for a moment any other reason for depositing them, such as the difficulties in re-hydrogenating the protein later on, etc.). So in the context of Phenix and Fcalc, do the positions of the riding hydrogens matter? And if they don't, does this mean that when an input model has zero hydrogens, Phenix is still trying in some way to account for the unmodeled/invisible hydrogens that we know are in there? Thanks, Damian On Nov 14, 2011, at 2:37 PM, Pavel Afonine wrote:
Hi Damian,
the contribution of H atoms to the total model structure factor Fmodel is refined in recent PHENIX versions, so these two parameters are not necessary and in fact are leftovers that I forgot to remove when switching to the new code. I will do it for the newer versions. Sorry for the confusion.
So, if you have H atoms in your input PDB file then you don't have to change any parameter. If you don't want to use H atoms then simply make sure they are not present in input PDB file.
Current implementation of handling of hydrogen atoms in refinement is partially inspired by J. Appl. Cryst. (2010). 43, 1100-1107 and thoroughly tested by various numerical tests (that we should probably publish in one of CCN issues).
Pavel
On 11/14/11 2:21 PM, Damian Ekiert wrote:
Hello,
I have a question regarding two parameters and how they affect riding hydrogens.
contribute_to_f_calc = True high_resolution_limit_to_include_scattering_from_h = 1.6
At lower resolution, setting contribute_to_f_calc = false is frequently helpful when using riding hydrogens and has been discussed before. But what is high_resolution_limit_to_include_scattering_from_h = 1.6 doing? If scattering from hydrogens is only included at higher than 1.6A resolution, it would seem like contribute_to_f_calc = True/False would not have any effect at say, 3.5A, but comparing identical refinement runs with contribute_to_f_calc = True or contribute_to_f_calc = False clearly gives different results. Can hydrogens contribute to Fcalc in a way aside from their scattering, or is their contribution accounted for independent of the position of the riding hydrogen (e.g., additional scattering term added to the associated heavy atom)? I think I am misunderstanding what these two parameters are doing and would appreciate some clarification.
Thanks,
Damian Ekiert _______________________________________________ phenixbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/phenixbb
_______________________________________________ phenixbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/phenixbb
On Mon, Nov 14, 2011 at 4:12 PM, Damian Ekiert
As I understand things, the positions of the riding hydrogens (from an input PDB file) would would be incorporated into the calculation of Fcalc (I guess automatically now and before when contribute_to_f_calc = True). If this is correct, then removing riding hydrogens before deposition seems very problematic, as they were effectively used in the calculation of R/R-free, etc. However, if the contribution of hydrogens is more abstract (e.g., the riding hydrogens are not used in the calculation, but the presence of hydrogens in the protein in the crystal are accounted for in some other way), leaving them out of the deposition is somewhat less of a problem (setting aside for a moment any other reason for depositing them, such as the difficulties in re-hydrogenating the protein later on, etc.).
For what it's worth, I've been doing a lot of data mining of new structures in the PDB recently, and the recalculated R-factors for high-resolution structures (including those originally refined with PHENIX) are typically 1-2% higher than reported. So stripping the hydrogens really is leaving out a non-trivial component of the model. (I don't see very much change at low resolution.)
So in the context of Phenix and Fcalc, do the positions of the riding hydrogens matter? And if they don't, does this mean that when an input model has zero hydrogens, Phenix is still trying in some way to account for the unmodeled/invisible hydrogens that we know are in there?
I think the answer to the first question is "sort of" (moreso for neutron than X-ray diffraction), but the answer to the second part is that it does account for them in the geometry restraints (by inflating atomic radii for the nonbonded interactions - the chirality restraints also compensate), but not in the scattering. (The scaling of F(calc) and F(obs) usually takes care of any resulting difference in R-factors at medium-to-low resolution, but I don't know how this changes the optimization target.) -Nat
Hi Damian,
This question was motivated by discussions of whether or not to deposit riding hydrogens in the PDB,
this was discussed so many times on both ccp4bb and phenixbb... So, very briefly, I do not know any single reason (and can't even imagine one) why one may want to remove hydrogens before PDB deposition. Most of reasons why hydrogen atoms should not be removed are listed in: phenix.model_vs_data: a high-level tool for the calculation of crystallographic model and data statistics. P.V. Afonine, R.W. Grosse-Kunstleve, V.B. Chen, J.J. Headd, N.W. Moriarty, J.S. Richardson, D.C. Richardson, A. Urzhumtsev, P.H. Zwart, P.D. Adams J. Appl. Cryst. 43, 677-685 (2010). and lots of hints here: Joint X-ray and neutron refinement with phenix.refine. P.V. Afonine, M. Mustyakimov, R.W. Grosse-Kunstleve, N.W. Moriarty, P. Langan, and P.D. Adams Acta Cryst. D66, 1153-1163 (2010).
So in the context of Phenix and Fcalc, do the positions of the riding hydrogens matter?
Yes, positions matter. As well as B-factors and occupancies. Note, B-factors of H are inherited from parent atoms with some multiplier that ranges from 1 to 1.5 (see SHELXL manual for details).
And if they don't, does this mean that when an input model has zero hydrogens, Phenix is still trying in some way to account for the unmodeled/invisible hydrogens that we know are in there?
No, if there are no H atoms in input PDB file, phenix.refine will not try to account for missing H in any way. Pavel
Sorry I'm chiming in a bit late here, but I think its worthwhile to point
out that hydrogen atoms are highly influenced by (and affect!) prior
chemical forces - primarily electrostatics/polarization and van der Waals.
Given the difficulty in refining hydrogen atoms against the scattering
data alone, I would argue a rigorous model for electrostatics/vdw terms in
the prior is essential. This also makes it a moot point of having to
decide how to refine hydrogen atoms and dealing with tricky cases (like
rotatable degrees of freedom or water molecules). There is technology
available to do this already and some publications on its importance in the
refinement process, for those that are interested.
Regards,
Tim
On Mon, Nov 14, 2011 at 6:27 PM, Pavel Afonine
** Hi Damian,
This question was motivated by discussions of whether or not to deposit riding hydrogens in the PDB,
this was discussed so many times on both ccp4bb and phenixbb... So, very briefly, I do not know any single reason (and can't even imagine one) why one may want to remove hydrogens before PDB deposition. Most of reasons why hydrogen atoms should not be removed are listed in:
phenix.model_vs_data: a high-level tool for the calculation of crystallographic model and data statistics. P.V. Afonine, R.W. Grosse-Kunstleve, V.B. Chen, J.J. Headd, N.W. Moriarty, J.S. Richardson, D.C. Richardson, A. Urzhumtsev, P.H. Zwart, P.D. Adams J. Appl. Cryst. 43, 677-685 (2010).
and lots of hints here:
Joint X-ray and neutron refinement with phenix.refine. P.V. Afonine, M. Mustyakimov, R.W. Grosse-Kunstleve, N.W. Moriarty, P. Langan, and P.D. Adams Acta Cryst. D66, 1153-1163 (2010).
So in the context of Phenix and Fcalc, do the positions of the riding hydrogens matter?
Yes, positions matter. As well as B-factors and occupancies. Note, B-factors of H are inherited from parent atoms with some multiplier that ranges from 1 to 1.5 (see SHELXL manual for details).
And if they don't, does this mean that when an input model has zero hydrogens, Phenix is still trying in some way to account for the unmodeled/invisible hydrogens that we know are in there?
No, if there are no H atoms in input PDB file, phenix.refine will not try to account for missing H in any way.
Pavel
_______________________________________________ phenixbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/phenixbb
participants (4)
-
Damian Ekiert -
Nathaniel Echols -
Pavel Afonine -
Tim Fenn