Hi I thought that a while ago, I read somewhere that rmsd(bonds) had not been calculated the same way in phenix as in refmac. I believe it was in connection with automated weights, and that rmsd(bonds) always much lower in phenix. I understood too that this had been adjusted since to match refmac. But we're still seeing some much lower rmsd(bond) from phenix: in particular, when we take a refmac model and plug it straight into phenix, the rmsd for cycle zero is already much lower than what it was at the end of refmac. Could someone enlighten...? (Yup, latest version of phenix.) Thanks! phx
Hi Frank,
But we're still seeing some much lower rmsd(bond) from phenix: in particular, when we take a refmac model and plug it straight into phenix, the rmsd for cycle zero is already much lower than what it was at the end of refmac.
Could someone enlighten...? (Yup, latest version of phenix.)
I have no idea about why they are different. I would approach this by taking a three-atoms test molecule, measure all bond lengths with a ruler (and calculate corresponding rmsd's) and then put this molecule into phenix.refine and refmac and see what they output. Do you have such example? Pavel.
PS. Also, can you get Refmac output all target values (ideal bond lengths, etc) used in restraints calculation? phenix.refine outputs a complete picture of restraints used (see .geo file). Is there anything similar in Refmac? If there is, then you could compare the values atom by atom. Pavel. On 12/15/2008 6:59 PM, Pavel Afonine wrote:
Hi Frank,
But we're still seeing some much lower rmsd(bond) from phenix: in particular, when we take a refmac model and plug it straight into phenix, the rmsd for cycle zero is already much lower than what it was at the end of refmac.
Could someone enlighten...? (Yup, latest version of phenix.)
I have no idea about why they are different. I would approach this by taking a three-atoms test molecule, measure all bond lengths with a ruler (and calculate corresponding rmsd's) and then put this molecule into phenix.refine and refmac and see what they output.
Do you have such example?
Pavel.
Hi, Has anyone solved this problem? I am also fighting with extremly low rmsd values. For 1.7A structure I get rmsd 0.004A (bond) 0.733 (angles) (R/Rfree=0.1884/0.2216) with automatic weight calculation. When I want to loose restraints to get more reasonable rmsd I get R/Rfree 0.1814/0.2240). However, at the moment I do not know what "reasonable" means while using phenix.refine. Is seems to me that the program include hydrogen atoms in rmsd calculations. I have done a test and run pdbtools for my structure refined with automatically adjusted weight and given manually. Below are the statistics given for a model with and without hydrogen atoms: Automatic with H Geometry statistics---------------------------------------------------------| | Type | Count | Deviation from ideal | Targets | Target (sum) | | | | rmsd max min | | | | bond | 5464 | 0.004 0.033 0.000 | 0.029 | | | angle | 9954 | 0.734 6.777 0.000 | 0.092 | |-----------------------------------------------------------------------------| Automatic without H |-Geometry statistics---------------------------------------------------------| | Type | Count | Deviation from ideal | Targets | Target (sum) | | | | rmsd max min | | | | bond | 2764 | 0.006 0.036 0.000 | 0.060 | | | angle | 3763 | 1.003 6.777 0.001 | 0.196 | | User-provided, with H -Geometry statistics---------------------------------------------------------| | Type | Count | Deviation from ideal | Targets | Target (sum) | | | | rmsd max min | | | | bond | 5464 | 0.023 0.177 0.000 | 0.924 | | | angle | 9954 | 1.592 12.640 0.000 | 0.549 | | User-provided, without H |-Geometry statistics---------------------------------------------------------| | Type | Count | Deviation from ideal | Targets | Target (sum) | | | | rmsd max min | | | | bond | 2764 | 0.032 0.177 0.000 | 1.832 | | | angle | 3763 | 2.343 12.640 0.001 | 1.317 | | Clearly, phenix.refine includes H in calculation. So if I want to get statistics similar to refmac output (which do not include H) I have to test several weights and check statistics given by pdbtools after H removal. Am I right? Karolina Dnia 2008-12-16, wto o godzinie 02:21 +0000, Frank von Delft pisze:
Hi
I thought that a while ago, I read somewhere that rmsd(bonds) had not been calculated the same way in phenix as in refmac. I believe it was in connection with automated weights, and that rmsd(bonds) always much lower in phenix.
I understood too that this had been adjusted since to match refmac.
But we're still seeing some much lower rmsd(bond) from phenix: in particular, when we take a refmac model and plug it straight into phenix, the rmsd for cycle zero is already much lower than what it was at the end of refmac.
Could someone enlighten...? (Yup, latest version of phenix.)
Thanks! phx _______________________________________________ phenixbb mailing list [email protected] http://www.phenix-online.org/mailman/listinfo/phenixbb
Hi Karolina,
Hi,
Has anyone solved this problem? I am also fighting with extremly low rmsd values. For 1.7A structure I get rmsd 0.004A (bond) 0.733 (angles) (R/Rfree=0.1884/0.2216) with automatic weight calculation. When I want to loose restraints to get more reasonable rmsd I get R/Rfree 0.1814/0.2240). However, at the moment I do not know what "reasonable" means while using phenix.refine. Is seems to me that the program include hydrogen atoms in rmsd calculations.
Put hydrogen atoms issue aside (see my previous email), and even put particular program use aside too, I would still be uncertain about "what reasonable means". There was enough of discussion on this matter in CCP4BB and relatively recent Acta D papers. I'm sure it can be easily criticized, but I would use that X-ray target weight that gives you the best Rfree (this is what "optimize_wxc=true" option in phenix.refine does, and similar for B-factors). And yes, for the moment you can use phenix.model_vs_data to report the final geometry rmsd statistics for a model containing hydrogen atoms: phenix.model_vs_data model.pdb data.mtz Please let me know if you have any questions! Cheers, Pavel.
Hi Pavel, It would be nice if the PDB would include phenix as one of the choices for reporting program used to calculate rmsd values in a deposition; at the moment we have to report as 'other'. Not your job I know - perhaps Paul could twist arms? Best, Mark -----Original Message----- From: Pavel Afonine [mailto:[email protected]] Sent: Sun 1/25/2009 1:02 AM To: PHENIX user mailing list Cc: Mariusz Jaskolski Subject: Re: [phenixbb] rmsd: phenix vs refmac Hi Karolina,
Hi,
Has anyone solved this problem? I am also fighting with extremly low rmsd values. For 1.7A structure I get rmsd 0.004A (bond) 0.733 (angles) (R/Rfree=0.1884/0.2216) with automatic weight calculation. When I want to loose restraints to get more reasonable rmsd I get R/Rfree 0.1814/0.2240). However, at the moment I do not know what "reasonable" means while using phenix.refine. Is seems to me that the program include hydrogen atoms in rmsd calculations.
Put hydrogen atoms issue aside (see my previous email), and even put particular program use aside too, I would still be uncertain about "what reasonable means". There was enough of discussion on this matter in CCP4BB and relatively recent Acta D papers. I'm sure it can be easily criticized, but I would use that X-ray target weight that gives you the best Rfree (this is what "optimize_wxc=true" option in phenix.refine does, and similar for B-factors). And yes, for the moment you can use phenix.model_vs_data to report the final geometry rmsd statistics for a model containing hydrogen atoms: phenix.model_vs_data model.pdb data.mtz Please let me know if you have any questions! Cheers, Pavel. _______________________________________________ phenixbb mailing list [email protected] http://www.phenix-online.org/mailman/listinfo/phenixbb
participants (5)
-
Frank von Delft -
Karolina Michalska -
Mayer, Mark (NIH/NICHD) [E] -
Pavel Afonine -
Pavel Afonine