Re: [ccp4bb] Calculating R-factor and maps from a Refmac model containing TLS downloaded from the PDB
On Wednesday 12 March 2008 18:30, Pavel Afonine wrote: > Hi Martin, > > > 2) in the context of PX, only the total "B factor" contribution to > > Fcalc needs to be positive definite, the TLS component might not be > > (though it is satisfying if it is) Correct (if I understand what you mean). Note that the PDB header may also contain an overall B correction that also contributes to the net individual ADPs. > Please correct me if I'm wrong My understanding was that the T and L > matrices must be positive definite, otherwise they do not have physical > sense. Not quite correct. If the tensors are NPD, then the rigid body assumption is violated. This does not necessarily mean that the description is nonsensical (although it could be). For instance, if you swing a normal wooden baseball bat[*] the motion can be described by TLS. Now imagine that the bat is made of rubber, and can bend as you swing it. In this case the tip of the bat will lag the grip and the body during the first part of your swing. This is still describable using TLS, but the L vector will go NPD. In the case of a protein stucture, such a model was probably not what the depositor intended. But I don't see an a priori reason to say that the description is invalid. The TLS model may correctly describe the distribution of individual ADPs in the structure, even if the physical interpretation diverges from a small set of truly rigid bodies. Of course, non-positive definite tensors may also be the result of a refinement that has simply gone bad. I'm not saying the model should be accepted without further inspection! Ethan Merritt [*] American readers: s/wooden/aluminum/ British readers:s/baseball/cricket/ > Yes, I understand that what's in the end important for the actual > calculations is the positive definiteness of the total B-factor (since > it goes as sqrt(det(B)) into denominator in electron density and > gradients calculation). > > > The PDB entries should contain the origin of the coordinate system > > to which the TLS parameters refer, and thus it is something you choose > > not something you calculate. > > > > OK, this partially the deviations I observe. Although, I'm still a bit > puzzled about why some differences are so large? Isn't it true that the > computed center of mass of a group should be pretty close to the > "chosen" one (at least for large groups)? > > Cheers, > Pavel. > > --- > Pavel V. Afonine, Ph.D. > Lawrence Berkeley National Lab, Berkeley CA, USA (http://www.lbl.gov/) > CCI: Computational Crystallography Initiative (http://cci.lbl.gov/) > PHENIX (http://phenix-online.org/) > -- Ethan A MerrittCourier Deliveries: 1959 NE Pacific Dept of Biochemistry Health Sciences Building University of Washington - Seattle WA 98195-7742
Re: [ccp4bb] Calculating R-factor and maps from a Refmac model containing TLS downloaded from the PDB
On Wed, 2008-03-12 at 18:30 -0700, Pavel Afonine wrote: > Hi Martin, > > > 2) in the context of PX, only the total "B factor" contribution to Fcalc > > needs to be positive definite, the TLS component might not be (though it is > > satisfying if it is) > > > > Please correct me if I'm wrong My understanding was that the T and L > matrices must be positive definite, otherwise they do not have physical > sense. Yes, I understand that what's in the end important for the actual > calculations is the positive definiteness of the total B-factor (since > it goes as sqrt(det(B)) into denominator in electron density and > gradients calculation). I suppose that is what I mean by "satisfying" ! If you have non-positive definite T and L, then you still have a valid mathematical model for the Fs, but the interpretation in terms of a simple rigid body model is flawed. This is not good, but I would say that you shouldn't take the rigid body model too literally even in the case of positive definite T and L. > > The PDB entries should contain the origin of the coordinate system to which > > the TLS parameters refer, and thus it is something you choose not something > > you calculate. > > > > OK, this partially the deviations I observe. Although, I'm still a bit > puzzled about why some differences are so large? Isn't it true that the > computed center of mass of a group should be pretty close to the > "chosen" one (at least for large groups)? I am unsure how you are computing the TLS origins. My point was that that is a choice, and you didn't say what your choice was. Refmac chooses the (occupancy-weighted) centroid. This should be pretty much the same as the centre of mass. And at least for large groups, it is likely to be close to the centre of reaction, but you may get larger discrepancies for smaller groups. Cheers Martyn > > Cheers, > Pavel. > > --- > Pavel V. Afonine, Ph.D. > Lawrence Berkeley National Lab, Berkeley CA, USA (http://www.lbl.gov/) > CCI: Computational Crystallography Initiative (http://cci.lbl.gov/) > PHENIX (http://phenix-online.org/) > > -- *** * * * Dr. Martyn Winn * * * * STFC Daresbury Laboratory, Daresbury, Warrington, WA4 4AD, U.K. * * Tel: +44 1925 603455E-mail: [EMAIL PROTECTED] * * Fax: +44 1925 603825Skype name: martyn.winn * * URL: http://www.ccp4.ac.uk/martyn/ * ***
Re: [ccp4bb] Calculating R-factor and maps from a Refmac model containing TLS downloaded from the PDB
Hi Dale, to my knowledge, sfcheck calculates the contribution of the bulk solvent using the double-exponential scaling model (Babinet model), but not using the mask model that REFMAC applies per default. I sometimes have to discuss similar R-factor differences with RCSB upon PDB submission of refined models ... Best regards, Dirk. Am 12.03.2008 um 22:20 schrieb Dale Tronrud: Hi, I am looking over a number of models from the PDB but have been unable to reproduce the R-factors for any model that was refined with Refmac and contains TLS parameters. I usually can't get within 5% of the reported value. On the other hand, I usually do pretty well for models w/o TLS. An example is the model 1nkz. The PDB header gives an R value of 17% but even when I use tlsanal in CCP4i to generate a PDB with anisotropic B's that mimic the TLS parameters I get an R value of 22.4% using SFCheck. (I'm not implying that I suspect any problem with 1nkz, in fact I have every reason to believe this is the great model its published stats indicate.) I've found a CCP4 BB letter that stated that SFCheck does not pay attention to anisotropic B's but that letter was dated 2002. I hope this limitation has been removed, or at least the output would mention this limitation. Setting up a refinement in Refmac involves a large overhead, since even for zero cycles of refinement the program insists on a complete stereochemical definition for the strange and wondrous groups in this model. I would just like to verify the R factor and calculate a proper map for inspection in Coot. Since I have many models I would like to look at, I would like a simple procedure. I did set up a Refmac run for another model, for which I do have all the .cif's required, but even after refinement I was not close to the reported R. I see that the models I'm interested in are not present in the Electron Density Server, so I suspect I'm not alone in fighting this battle. Any advice would be appreciated, Dale Tronrud *** Dirk Kostrewa Gene Center, A 5.07 Ludwig-Maximilians-University Feodor-Lynen-Str. 25 81377 Munich Germany Phone: +49-89-2180-76845 Fax:+49-89-2180-76999 E-mail: [EMAIL PROTECTED] ***
Re: [ccp4bb] Calculating R-factor and maps from a Refmac model containing TLS downloaded from the PDB
Dale, I have had a similar problem with two of my entries. However, if I take the TLS information from the PDB header and put it in a "in_tls.tls" file that I feed into REFMAC (0 cycles, just structure factor computation), I can reproduce the R-factors within 0.1% or less. It is just the order of the numbers that matters but this is described in the REFMAC documentation. Cheers, Ulrich On Wed, 12 Mar 2008 22:20:58 +0100, Dale Tronrud <[EMAIL PROTECTED]> wrote: Hi, I am looking over a number of models from the PDB but have been unable to reproduce the R-factors for any model that was refined with Refmac and contains TLS parameters. I usually can't get within 5% of the reported value. On the other hand, I usually do pretty well for models w/o TLS. An example is the model 1nkz. The PDB header gives an R value of 17% but even when I use tlsanal in CCP4i to generate a PDB with anisotropic B's that mimic the TLS parameters I get an R value of 22.4% using SFCheck. (I'm not implying that I suspect any problem with 1nkz, in fact I have every reason to believe this is the great model its published stats indicate.) I've found a CCP4 BB letter that stated that SFCheck does not pay attention to anisotropic B's but that letter was dated 2002. I hope this limitation has been removed, or at least the output would mention this limitation. Setting up a refinement in Refmac involves a large overhead, since even for zero cycles of refinement the program insists on a complete stereochemical definition for the strange and wondrous groups in this model. I would just like to verify the R factor and calculate a proper map for inspection in Coot. Since I have many models I would like to look at, I would like a simple procedure. I did set up a Refmac run for another model, for which I do have all the .cif's required, but even after refinement I was not close to the reported R. I see that the models I'm interested in are not present in the Electron Density Server, so I suspect I'm not alone in fighting this battle. Any advice would be appreciated, Dale Tronrud -- Prof Ulrich Baumann Departement für Chemie und Biochemie Freiestrasse 3 Ch-3012 Bern, Switzerland phone + 41 (0)31 631 4320/4343 fax + 41 (0) 31 631 4887
Re: [ccp4bb] Calculating R-factor and maps from a Refmac model containing TLS downloaded from the PDB
Hi Martin, 2) in the context of PX, only the total "B factor" contribution to Fcalc needs to be positive definite, the TLS component might not be (though it is satisfying if it is) Please correct me if I'm wrong My understanding was that the T and L matrices must be positive definite, otherwise they do not have physical sense. Yes, I understand that what's in the end important for the actual calculations is the positive definiteness of the total B-factor (since it goes as sqrt(det(B)) into denominator in electron density and gradients calculation). The PDB entries should contain the origin of the coordinate system to which the TLS parameters refer, and thus it is something you choose not something you calculate. OK, this partially the deviations I observe. Although, I'm still a bit puzzled about why some differences are so large? Isn't it true that the computed center of mass of a group should be pretty close to the "chosen" one (at least for large groups)? Cheers, Pavel. --- Pavel V. Afonine, Ph.D. Lawrence Berkeley National Lab, Berkeley CA, USA (http://www.lbl.gov/) CCI: Computational Crystallography Initiative (http://cci.lbl.gov/) PHENIX (http://phenix-online.org/)
Re: [ccp4bb] Calculating R-factor and maps from a Refmac model containing TLS downloaded from the PDB
Clearly a few interesting entries . 2) in the context of PX, only the total "B factor" contribution to Fcalc needs to be positive definite, the TLS component might not be (though it is satisfying if it is) 3) how do you calculate the "TLS origin" ? The PDB entries should contain the origin of the coordinate system to which the TLS parameters refer, and thus it is something you choose not something you calculate. You are perhaps comparing the origin used by refmac with the centre of reaction ?? Cheers Martyn -Original Message- From: CCP4 bulletin board on behalf of Pavel Afonine Sent: Wed 3/12/2008 10:48 PM To: CCP4BB@JISCMAIL.AC.UK Subject: Re: [ccp4bb] Calculating R-factor and maps from a Refmac model containing TLS downloaded from the PDB Hi Dale, a few of my observations regarding this issue: 1) A quick run over all TLS containing models in PDB results in this list of models with incorrect TLS selections or other artifacts (see "errors1" file, attached). 2) A lot of reported TLS matrices in PDB are not positive definite. You can easily verify it (I did). 3) Reported TLS origins are not always the same if you compute them yourself (see "errors2" file, attached). This file shows the distance between reported and re-computed TLS origin for each group. 4) Some files contain residual B-factors (Btotal-Btls), and some files contain total B-factors (Bresidual + Btls). ... I have a *very* long list. I re-refine all PDB models (for which data are available) from time to time -:). Let me know if interested. And finally, how you like the anisotropic B-factors in this file -:) http://www.rcsb.org/pdb/files/2atb.pdb Without some prior re-refinement you have almost no chances to reproduce the R-factors for most of these structures. Cheers, Pavel. --- Pavel V. Afonine, Ph.D. Lawrence Berkeley National Lab, Berkeley CA, USA (http://www.lbl.gov/) CCI: Computational Crystallography Initiative (http://cci.lbl.gov/) PHENIX (http://phenix-online.org/) On 3/12/2008 2:20 PM, Dale Tronrud wrote: > Hi, > >I am looking over a number of models from the PDB but have been > unable to reproduce the R-factors for any model that was refined > with Refmac and contains TLS parameters. I usually can't get within > 5% of the reported value. On the other hand, I usually do pretty > well for models w/o TLS. > >An example is the model 1nkz. The PDB header gives an R value > of 17% but even when I use tlsanal in CCP4i to generate a PDB with > anisotropic B's that mimic the TLS parameters I get an R value of > 22.4% using SFCheck. (I'm not implying that I suspect any problem > with 1nkz, in fact I have every reason to believe this is the great > model its published stats indicate.) > >I've found a CCP4 BB letter that stated that SFCheck does not > pay attention to anisotropic B's but that letter was dated 2002. > I hope this limitation has been removed, or at least the output > would mention this limitation. > >Setting up a refinement in Refmac involves a large overhead, > since even for zero cycles of refinement the program insists on > a complete stereochemical definition for the strange and wondrous > groups in this model. I would just like to verify the R factor > and calculate a proper map for inspection in Coot. Since I have > many models I would like to look at, I would like a simple procedure. > >I did set up a Refmac run for another model, for which I do > have all the .cif's required, but even after refinement I was not > close to the reported R. > >I see that the models I'm interested in are not present in the > Electron Density Server, so I suspect I'm not alone in fighting > this battle. > > Any advice would be appreciated, > Dale Tronrud
Re: [ccp4bb] Calculating R-factor and maps from a Refmac model containing TLS downloaded from the PDB
Dale, You raise a very important point. The electron density server will not put out the data unless there is a reasonable agreement between the Author reported R-factor and the calculated R-factor. I guess for the same reasons (and more - Gerard?). The question I have is - if we are going to ask people to deposit data with their model - what information should be deposited (and be in the header of the CIF or PDB file), that would allow us to reproduce the exact maps and statistics that the author reports? We have been struggling with that for some time now in my lab. If the community can come up with an answer, then I am sure we can convince PDB to add those information. Best, Rams. S. Ramaswamy Department of Biochemistry University of Iowa. On Wed, 2008-03-12 at 14:20 -0700, Dale Tronrud wrote: > Hi, > > I am looking over a number of models from the PDB but have been > unable to reproduce the R-factors for any model that was refined > with Refmac and contains TLS parameters. I usually can't get within > 5% of the reported value. On the other hand, I usually do pretty > well for models w/o TLS. > > An example is the model 1nkz. The PDB header gives an R value > of 17% but even when I use tlsanal in CCP4i to generate a PDB with > anisotropic B's that mimic the TLS parameters I get an R value of > 22.4% using SFCheck. (I'm not implying that I suspect any problem > with 1nkz, in fact I have every reason to believe this is the great > model its published stats indicate.) > > I've found a CCP4 BB letter that stated that SFCheck does not > pay attention to anisotropic B's but that letter was dated 2002. > I hope this limitation has been removed, or at least the output > would mention this limitation. > > Setting up a refinement in Refmac involves a large overhead, > since even for zero cycles of refinement the program insists on > a complete stereochemical definition for the strange and wondrous > groups in this model. I would just like to verify the R factor > and calculate a proper map for inspection in Coot. Since I have > many models I would like to look at, I would like a simple procedure. > > I did set up a Refmac run for another model, for which I do > have all the .cif's required, but even after refinement I was not > close to the reported R. > > I see that the models I'm interested in are not present in the > Electron Density Server, so I suspect I'm not alone in fighting > this battle. > > Any advice would be appreciated, > Dale Tronrud
[ccp4bb] Calculating R-factor and maps from a Refmac model containing TLS downloaded from the PDB
Hi, I am looking over a number of models from the PDB but have been unable to reproduce the R-factors for any model that was refined with Refmac and contains TLS parameters. I usually can't get within 5% of the reported value. On the other hand, I usually do pretty well for models w/o TLS. An example is the model 1nkz. The PDB header gives an R value of 17% but even when I use tlsanal in CCP4i to generate a PDB with anisotropic B's that mimic the TLS parameters I get an R value of 22.4% using SFCheck. (I'm not implying that I suspect any problem with 1nkz, in fact I have every reason to believe this is the great model its published stats indicate.) I've found a CCP4 BB letter that stated that SFCheck does not pay attention to anisotropic B's but that letter was dated 2002. I hope this limitation has been removed, or at least the output would mention this limitation. Setting up a refinement in Refmac involves a large overhead, since even for zero cycles of refinement the program insists on a complete stereochemical definition for the strange and wondrous groups in this model. I would just like to verify the R factor and calculate a proper map for inspection in Coot. Since I have many models I would like to look at, I would like a simple procedure. I did set up a Refmac run for another model, for which I do have all the .cif's required, but even after refinement I was not close to the reported R. I see that the models I'm interested in are not present in the Electron Density Server, so I suspect I'm not alone in fighting this battle. Any advice would be appreciated, Dale Tronrud