Hello, I have a good data set to 1.8 A and refined it with the actual version of phenix, including annealing and isotropic bfactors. Checking the B factor variance in coot revealed high differences in about 50% of the residues. Most of them are at the surface, but still there are residues like phenylalanines and valines within the core where the bfactors jump from 25 to 50 from one carbon atom to the other. Changing wu and wxu_scale did not improve the results. A quick check with refmac improved the results, so i guess I am missing the right restraints for phenix. Which parameters should be changed to improve the bfactor variance in phenix? Thanks for your help Daniel Daniel Frey Department of Biochmistry University of Zurich Winterthurerstrass 190 8057 Zurich Switzerland [email protected] Tel: +41446355558 www.biochem.uzh.ch/gruetter www.structuralbiology.uzh.ch
Daniel, this appears to be a common thread on this BB, since phenix seems to handle these things quite differently than CNS and refmac(?). A while back I had a similar exchange with Pavel about this. He pointed me to some literature which highlights some of the challenges associated with restrained B-factor refinement using medium to low resolution data. From what I understand the implementation in PHENIX is based on the ideas of Ian Tickle. Here is also a paper I found useful: http://www.phenix-online.org/papers/ccp4_july_2005_afonine.pdf That should give you some pointers were to look for tweakable parameters. HTH Carsten ------------------------------------------------------------------------------------------
Does anyone know a publication which covers theoretical B-factor distributions?
Here are a few links related to your question that came to my mind first: D. E. Tronrud J. Appl. Cryst. (1996). 29, 100-104 Knowledge-Based B-Factor Restraints for the Refinement of Proteins F. L. Hirshfeld Acta Cryst. (1976). A32, 239-244 Can X-ray data distinguish bonding effects from vibrational smearing? S. PARTHASARATHY AND M.R.N. MURTHY Analysis of temperature factor distribution in high-resolution protein structures Protein Science (1997), 6:2561-2567. http://www.ysbl.york.ac.uk/ccp4bb/2003/msg00432.html http://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=ind0801&L=ccp4bb&T=0&F=&S=&P=52510 J.D. DUNITZ and D.N.J. WHITE Non-rigid-body thermal-motion analysis. Acta Cryst. (1973). A29, 93 Cheers, Pavel. ------------------------------------------------------------------------------------------
-----Original Message----- From: [email protected] [mailto:[email protected]]On Behalf Of Daniel Frey Sent: Thursday, March 20, 2008 7:00 AM To: [email protected] Subject: [phenixbb] ADP high variance
Hello, I have a good data set to 1.8 A and refined it with the actual version of phenix, including annealing and isotropic bfactors. Checking the B factor variance in coot revealed high differences in about 50% of the residues. Most of them are at the surface, but still there are residues like phenylalanines and valines within the core where the bfactors jump from 25 to 50 from one carbon atom to the other. Changing wu and wxu_scale did not improve the results. A quick check with refmac improved the results, so i guess I am missing the right restraints for phenix. Which parameters should be changed to improve the bfactor variance in phenix?
Thanks for your help
Daniel
Daniel Frey Department of Biochmistry University of Zurich Winterthurerstrass 190 8057 Zurich Switzerland
[email protected] Tel: +41446355558
www.biochem.uzh.ch/gruetter www.structuralbiology.uzh.ch
_______________________________________________ phenixbb mailing list [email protected] http://www.phenix-online.org/mailman/listinfo/phenixbb
Hi Daniel, Carsten is right, phenix.refine uses a different approach. However, changing the wxu_scale should visibly change the ADP variances. - are you using TLS? - what is the exact command you run? - if your PDB file has segment identifiers, are they all consistent (no mixes)? Normally, changing the wxu_scale has to change the ADP variances. If it does not change then something is not right somewhere. I need some more information to tell you exactly what's happening. If you want me to debug it some more, I'm happy to do so but I will need to reproduce this problem on myself on computer (meaning I need the model and data). Thanks! Pavel. Daniel Frey wrote:
Hello, I have a good data set to 1.8 A and refined it with the actual version of phenix, including annealing and isotropic bfactors. Checking the B factor variance in coot revealed high differences in about 50% of the residues. Most of them are at the surface, but still there are residues like phenylalanines and valines within the core where the bfactors jump from 25 to 50 from one carbon atom to the other. Changing wu and wxu_scale did not improve the results. A quick check with refmac improved the results, so i guess I am missing the right restraints for phenix. Which parameters should be changed to improve the bfactor variance in phenix?
Thanks for your help
Daniel
Daniel Frey Department of Biochmistry University of Zurich Winterthurerstrass 190 8057 Zurich Switzerland
[email protected] Tel: +41446355558
www.biochem.uzh.ch/gruetter www.structuralbiology.uzh.ch
_______________________________________________ phenixbb mailing list [email protected] http://www.phenix-online.org/mailman/listinfo/phenixbb
Dear Pavel and Carsten, I finally managed to change the ADP variance by setting the wxu_scale below 0.02. Thanks for the advise Daniel On Mar 20, 2008, at 3:43 PM, Pavel Afonine wrote:
Hi Daniel,
Carsten is right, phenix.refine uses a different approach. However, changing the wxu_scale should visibly change the ADP variances.
- are you using TLS? - what is the exact command you run? - if your PDB file has segment identifiers, are they all consistent (no mixes)?
Normally, changing the wxu_scale has to change the ADP variances. If it does not change then something is not right somewhere. I need some more information to tell you exactly what's happening. If you want me to debug it some more, I'm happy to do so but I will need to reproduce this problem on myself on computer (meaning I need the model and data).
Thanks! Pavel.
Daniel Frey wrote:
Hello, I have a good data set to 1.8 A and refined it with the actual version of phenix, including annealing and isotropic bfactors. Checking the B factor variance in coot revealed high differences in about 50% of the residues. Most of them are at the surface, but still there are residues like phenylalanines and valines within the core where the bfactors jump from 25 to 50 from one carbon atom to the other. Changing wu and wxu_scale did not improve the results. A quick check with refmac improved the results, so i guess I am missing the right restraints for phenix. Which parameters should be changed to improve the bfactor variance in phenix?
Thanks for your help
Daniel
Daniel Frey Department of Biochmistry University of Zurich Winterthurerstrass 190 8057 Zurich Switzerland
[email protected] Tel: +41446355558
www.biochem.uzh.ch/gruetter www.structuralbiology.uzh.ch
_______________________________________________ phenixbb mailing list [email protected] http://www.phenix-online.org/mailman/listinfo/phenixbb
_______________________________________________ phenixbb mailing list [email protected] http://www.phenix-online.org/mailman/listinfo/phenixbb
Daniel Frey Department of Biochemistry University of Zurich Winterthurerstrasse 190 8057 Zurich Switzerland [email protected] Tel: +41446355558 www.biochem.uzh.ch/gruetter www.structuralbiology.uzh.ch
Daniel, What was the effect on your R/Rfree?
-----Original Message----- From: [email protected] [mailto:[email protected]]On Behalf Of Daniel Frey Sent: Tuesday, March 25, 2008 1:15 PM To: PHENIX user mailing list Subject: Re: [phenixbb] ADP high variance
Dear Pavel and Carsten,
I finally managed to change the ADP variance by setting the wxu_scale below 0.02. Thanks for the advise
Daniel
On Mar 20, 2008, at 3:43 PM, Pavel Afonine wrote:
Hi Daniel,
Carsten is right, phenix.refine uses a different approach. However, changing the wxu_scale should visibly change the ADP variances.
- are you using TLS? - what is the exact command you run? - if your PDB file has segment identifiers, are they all consistent (no mixes)?
Normally, changing the wxu_scale has to change the ADP variances. If it does not change then something is not right somewhere. I need some more information to tell you exactly what's happening. If you want me to debug it some more, I'm happy to do so but I will need to reproduce this problem on myself on computer (meaning I need the model and data).
Thanks! Pavel.
Daniel Frey wrote:
Hello, I have a good data set to 1.8 A and refined it with the actual version of phenix, including annealing and isotropic bfactors. Checking the B factor variance in coot revealed high differences in about 50% of the residues. Most of them are at the surface, but still there are residues like phenylalanines and valines within the core where the bfactors jump from 25 to 50 from one carbon atom to the other. Changing wu and wxu_scale did not improve the results. A quick check with refmac improved the results, so i guess I am missing the right restraints for phenix. Which parameters should be changed to improve the bfactor variance in phenix?
Thanks for your help
Daniel
Daniel Frey Department of Biochmistry University of Zurich Winterthurerstrass 190 8057 Zurich Switzerland
[email protected] Tel: +41446355558
www.biochem.uzh.ch/gruetter www.structuralbiology.uzh.ch
_______________________________________________ phenixbb mailing list [email protected] http://www.phenix-online.org/mailman/listinfo/phenixbb
_______________________________________________ phenixbb mailing list [email protected] http://www.phenix-online.org/mailman/listinfo/phenixbb
Daniel Frey Department of Biochemistry University of Zurich Winterthurerstrasse 190 8057 Zurich Switzerland
[email protected] Tel: +41446355558
www.biochem.uzh.ch/gruetter www.structuralbiology.uzh.ch
_______________________________________________ phenixbb mailing list [email protected] http://www.phenix-online.org/mailman/listinfo/phenixbb
Dear Carsten, the model with the high variance had R and Rfree of 17.8 and 22.2 (model1) after switching to wxu_scale 0.01 it was 17.9 and 21.7 (model2) but still the adp variance was unacceptable high. For the next run I set the initial Bfactor of the model to 26 and rerun the script, R and Rfree went up to 20.5 and 25.5. Here all ADPs were perfect and outliners were not detected in Coot. Now I am trying to find the optimal weight for wxu_scale, just tried to go again higher with the wxu_scale and at 0.1 the adp variance still looks fine and the R values dropped to 18.6 and 22.8. Greetings Daniel On Mar 25, 2008, at 6:38 PM, Schubert, Carsten [PRDUS] wrote:
Daniel,
What was the effect on your R/Rfree?
-----Original Message----- From: [email protected] [mailto:[email protected]]On Behalf Of Daniel Frey Sent: Tuesday, March 25, 2008 1:15 PM To: PHENIX user mailing list Subject: Re: [phenixbb] ADP high variance
Dear Pavel and Carsten,
I finally managed to change the ADP variance by setting the wxu_scale below 0.02. Thanks for the advise
Daniel
On Mar 20, 2008, at 3:43 PM, Pavel Afonine wrote:
Hi Daniel,
Carsten is right, phenix.refine uses a different approach. However, changing the wxu_scale should visibly change the ADP variances.
- are you using TLS? - what is the exact command you run? - if your PDB file has segment identifiers, are they all consistent (no mixes)?
Normally, changing the wxu_scale has to change the ADP variances. If it does not change then something is not right somewhere. I need some more information to tell you exactly what's happening. If you want me to debug it some more, I'm happy to do so but I will need to reproduce this problem on myself on computer (meaning I need the model and data).
Thanks! Pavel.
Daniel Frey wrote:
Hello, I have a good data set to 1.8 A and refined it with the actual version of phenix, including annealing and isotropic bfactors. Checking the B factor variance in coot revealed high differences in about 50% of the residues. Most of them are at the surface, but still there are residues like phenylalanines and valines within the core where the bfactors jump from 25 to 50 from one carbon atom to the other. Changing wu and wxu_scale did not improve the results. A quick check with refmac improved the results, so i guess I am missing the right restraints for phenix. Which parameters should be changed to improve the bfactor variance in phenix?
Thanks for your help
Daniel
Daniel Frey Department of Biochmistry University of Zurich Winterthurerstrass 190 8057 Zurich Switzerland
[email protected] Tel: +41446355558
www.biochem.uzh.ch/gruetter www.structuralbiology.uzh.ch
_______________________________________________ phenixbb mailing list [email protected] http://www.phenix-online.org/mailman/listinfo/phenixbb
_______________________________________________ phenixbb mailing list [email protected] http://www.phenix-online.org/mailman/listinfo/phenixbb
Daniel Frey Department of Biochemistry University of Zurich Winterthurerstrasse 190 8057 Zurich Switzerland
[email protected] Tel: +41446355558
www.biochem.uzh.ch/gruetter www.structuralbiology.uzh.ch
_______________________________________________ phenixbb mailing list [email protected] http://www.phenix-online.org/mailman/listinfo/phenixbb
_______________________________________________ phenixbb mailing list [email protected] http://www.phenix-online.org/mailman/listinfo/phenixbb
Daniel Frey Department of Biochemistry University of Zurich Winterthurerstrasse 190 8057 Zurich Switzerland [email protected] Tel: +41446355558 www.biochem.uzh.ch/gruetter www.structuralbiology.uzh.ch
Hi Daniel, I'm glad you managed this to work the way you like. However, the wxu_scale value looks very small (especially at 1.8A resolution) and this fact makes me think that the root of the problem is something else. Cheers, Pavel. Daniel Frey wrote:
Dear Pavel and Carsten,
I finally managed to change the ADP variance by setting the wxu_scale below 0.02. Thanks for the advise
Daniel
On Mar 20, 2008, at 3:43 PM, Pavel Afonine wrote:
Hi Daniel,
Carsten is right, phenix.refine uses a different approach. However, changing the wxu_scale should visibly change the ADP variances.
- are you using TLS? - what is the exact command you run? - if your PDB file has segment identifiers, are they all consistent (no mixes)?
Normally, changing the wxu_scale has to change the ADP variances. If it does not change then something is not right somewhere. I need some more information to tell you exactly what's happening. If you want me to debug it some more, I'm happy to do so but I will need to reproduce this problem on myself on computer (meaning I need the model and data).
Thanks! Pavel.
Daniel Frey wrote:
Hello, I have a good data set to 1.8 A and refined it with the actual version of phenix, including annealing and isotropic bfactors. Checking the B factor variance in coot revealed high differences in about 50% of the residues. Most of them are at the surface, but still there are residues like phenylalanines and valines within the core where the bfactors jump from 25 to 50 from one carbon atom to the other. Changing wu and wxu_scale did not improve the results. A quick check with refmac improved the results, so i guess I am missing the right restraints for phenix. Which parameters should be changed to improve the bfactor variance in phenix?
Thanks for your help
Daniel
Daniel Frey Department of Biochmistry University of Zurich Winterthurerstrass 190 8057 Zurich Switzerland
[email protected] Tel: +41446355558
www.biochem.uzh.ch/gruetter www.structuralbiology.uzh.ch
_______________________________________________ phenixbb mailing list [email protected] http://www.phenix-online.org/mailman/listinfo/phenixbb
_______________________________________________ phenixbb mailing list [email protected] http://www.phenix-online.org/mailman/listinfo/phenixbb
Daniel Frey Department of Biochemistry University of Zurich Winterthurerstrasse 190 8057 Zurich Switzerland
[email protected] Tel: +41446355558
www.biochem.uzh.ch/gruetter www.structuralbiology.uzh.ch
_______________________________________________ phenixbb mailing list [email protected] http://www.phenix-online.org/mailman/listinfo/phenixbb
participants (3)
-
Daniel Frey -
Pavel Afonine -
Schubert, Carsten [PRDUS]