Dear helpers in phenix, I have anomalous data (Co; wavelength: 1.609A) that I want to use its Phase information from SAD-phasing method (autosol). As I have a native data (from nomal wavelength ~1A) with higher resolution (~2A) than SAD data (~2.5A), I wanted to apply phase information to this native data for further refinement: My goals were 1. I want to confirm where is my heavy metal is; 2. I want to refine further using native data as has higher resolution. At first, when I did autosol, it didn't give me perfect result. So far what I tired is as following: 'SAD peak' from imosflm generated mtz file (scaled), 'Phases' from first refmac generated mtz file from the native data, and 'MIR native' from the native data mtz file (imosflm and scaled). FOM: 0.186; overall score was 10.75 +/- 12.33. FYI. I input 2 sites and result suggested 3 sites. Each molecule has one binding site of metal ion. I was using old Mac version:1.6.4-486 Anyway, I simply passed the result over Autobuild. It ran and gave me better statistics as following: R/Rfree=0.2734/0.3312; CC 0.76. However (I have two molecules/ASU) one of models was not great. But this time, anomalous density map well phased(?!) to 2Fo-Fc map and I could see - it was before only one heavy metal I clearly could see- 2 heavy metal signal from 2 molecules and those positions are well matched with native density maps. So I continued phenix.refine using Autobuild made 'expt_fobs_phase_freeR_flags_1.mtz'. The question is the statistics from anomalous data itself (it has slight dispcrepancy between anomalous map and 2Fo-Fc/Fo-Fc map due to phase problem) or even those from native data is better than my current model which came from native data set (SAD-phased as described above) with higher resolution. I think one of main reasons is one of molecule has a lid that does not perfectly fit into current density map. However, in terms of heavy metal position and the active site is better from current map (Autobuild generated model). FYI: current model has R/Rfree=0.21/0.27(best was 0.21/0.26); 0.19/0.24 (for native data only used model) .... Am I too concerning for statistics or is there something I am missing? Now my questions are as following plus any suggestions are welcome: 1. For SAD-phasing, could you tell me which is best way to do so with my case (one data from the edge of heavy metal (lower resolution) and native data (higher resolution) or where is the point I can convince. Also I tried to use DANO/SIGDANO column (from hkl2000 and ccp4) but seems like it (Autosol) doesn't take them into account. Could you tell me how can I improve phase by using anomalous data? 2. In case of Co, does anyone have experience about coordination numbers? I thought it should have 6 (or 4) coordination numbers. The question what I have now is about the numbers and their lengths ... I though it should be within 2A, but not true for my current models (And somehow distorted octahedron or else with 2.07-3.28A). I also searched PDB files, and found that many deviations from 1.98 to 2.82A (depending on corresponding atoms and resolutions: 3KOI, 3KON, 2DD5, and 2XJM). I would appreciate if you can give me any tips regarding this issue. Many thanks in advance, Young-Jin
Dear Young Jin, you should first aim on getting good density for the second (I guess identical to the first) molecule using your anomalous data before you switch to final model refinement with your native dataset. There are several methods to achieve this. You can for example: (1) use your (experimental) phases to facilitate molecular replacement using your well fitted model. This is possible in MOLREP (option: phased molecular replacement) and I guess also in phenix.automr (?). (2) use the well fitted model as a fixed model, delete the badly fitting one and use an additional first model as a search model (both PHASER and MOLREP) (3) use your well fitted model and eventually parts of the second one as a starting model for phenix.autosol (MR-SAD) to improve the exp. phases. Try different combinations of these methods and see what helps. In addition, assuming you have good quality data you should be able to see peaks from ordered Met/Cys sulfurs in an anomalous difference map as soon as you have a reasonable model. You can use this information to fix your model before you switch to final refinement with your native data. Good luck, Florian On 10.12.10 18:47, YoungJin wrote:
Dear helpers in phenix,
I have anomalous data (Co; wavelength: 1.609A) that I want to use its Phase information from SAD-phasing method (autosol). As I have a native data (from nomal wavelength ~1A) with higher resolution (~2A) than SAD data (~2.5A), I wanted to apply phase information to this native data for further refinement: My goals were 1. I want to confirm where is my heavy metal is; 2. I want to refine further using native data as has higher resolution. At first, when I did autosol, it didn't give me perfect result. So far what I tired is as following: 'SAD peak' from imosflm generated mtz file (scaled), 'Phases' from first refmac generated mtz file from the native data, and 'MIR native' from the native data mtz file (imosflm and scaled). FOM: 0.186; overall score was 10.75 +/- 12.33. FYI. I input 2 sites and result suggested 3 sites. Each molecule has one binding site of metal ion. I was using old Mac version:1.6.4-486
Anyway, I simply passed the result over Autobuild. It ran and gave me better statistics as following: R/Rfree=0.2734/0.3312; CC 0.76. However (I have two molecules/ASU) one of models was not great. But this time, anomalous density map well phased(?!) to 2Fo-Fc map and I could see - it was before only one heavy metal I clearly could see- 2 heavy metal signal from 2 molecules and those positions are well matched with native density maps.
So I continued phenix.refine using Autobuild made 'expt_fobs_phase_freeR_flags_1.mtz'. The question is the statistics from anomalous data itself (it has slight dispcrepancy between anomalous map and 2Fo-Fc/Fo-Fc map due to phase problem) or even those from native data is better than my current model which came from native data set (SAD-phased as described above) with higher resolution. I think one of main reasons is one of molecule has a lid that does not perfectly fit into current density map. However, in terms of heavy metal position and the active site is better from current map (Autobuild generated model). FYI: current model has R/Rfree=0.21/0.27(best was 0.21/0.26); 0.19/0.24 (for native data only used model) .... Am I too concerning for statistics or is there something I am missing?
Now my questions are as following plus any suggestions are welcome: 1. For SAD-phasing, could you tell me which is best way to do so with my case (one data from the edge of heavy metal (lower resolution) and native data (higher resolution) or where is the point I can convince. Also I tried to use DANO/SIGDANO column (from hkl2000 and ccp4) but seems like it (Autosol) doesn't take them into account. Could you tell me how can I improve phase by using anomalous data?
2. In case of Co, does anyone have experience about coordination numbers? I thought it should have 6 (or 4) coordination numbers. The question what I have now is about the numbers and their lengths ... I though it should be within 2A, but not true for my current models (And somehow distorted octahedron or else with 2.07-3.28A). I also searched PDB files, and found that many deviations from 1.98 to 2.82A (depending on corresponding atoms and resolutions: 3KOI, 3KON, 2DD5, and 2XJM). I would appreciate if you can give me any tips regarding this issue.
Many thanks in advance,
Young-Jin _______________________________________________ phenixbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/phenixbb
participants (2)
-
Florian Sauer -
YoungJin