Re: [ccp4bb] PHENIX BSS refinement
Hi Miguel, Right, but why this happens only in the final bss step ? Why not in the first one? It's arbitrary: one can do it every macro-cycle as well. The main idea is to have a total B-factor in ATOM records written out to a PDB file. I have noticed the behaviour described by José when I refine in phenix a model previously refined in refmac5 _with_ TLS, so the ADPs in the model are actually 'residual' ADPs of refmac5 TLS refinement. In these cases, during the refinement in phenix, the ADPs seem to be kept at very low values all throughout until they abruptly go to the higher, more reasonable values, after the final bulk solvent correction step. There may be a good reason for this behaviour, but I don't see it. I would appreciate if you could elaborate. Like I said, it is arbitrary. You can store the overall B-factor in overall anisotropic scale matrix (B_overall) or in individual atomic B-factors; in both cases the total model structure factor Fmodel = scale_overall * exp(-h*B_overall*ht) * (Fcalc + k_sol * exp(-B_sol*s^2) * Fmask) will remain the same. Pavel.
Re: [ccp4bb] PHENIX BSS refinement
Hi Pavel, Thanks, it is clear now. I have a last concern, though. Does phenix impose minimal values to the refined ADPs? I guess it does not, for I have found nothing like that in the documentation. But, if it does so, are they imposed to the individual B-factors or the B-overall is also taken into account? In the former case you might have problems with the refinement in a case as the one I described in my previous e-mail (coming from a model with artificially low, 'residual' B-factors). Cheers, Miguel Le 4 févr. 09 à 18:23, Pavel Afonine a écrit : Hi Miguel, Right, but why this happens only in the final bss step ? Why not in the first one? It's arbitrary: one can do it every macro-cycle as well. The main idea is to have a total B-factor in ATOM records written out to a PDB file. I have noticed the behaviour described by José when I refine in phenix a model previously refined in refmac5 _with_ TLS, so the ADPs in the model are actually 'residual' ADPs of refmac5 TLS refinement. In these cases, during the refinement in phenix, the ADPs seem to be kept at very low values all throughout until they abruptly go to the higher, more reasonable values, after the final bulk solvent correction step. There may be a good reason for this behaviour, but I don't see it. I would appreciate if you could elaborate. Like I said, it is arbitrary. You can store the overall B-factor in overall anisotropic scale matrix (B_overall) or in individual atomic B-factors; in both cases the total model structure factor Fmodel = scale_overall * exp(-h*B_overall*ht) * (Fcalc + k_sol * exp(-B_sol*s^2) * Fmask) will remain the same. Pavel. -- This message has been scanned for viruses and dangerous content by MailScanner, and is believed to be clean. -- Miguel Ortiz Lombardía Architecture et Fonction des Macromolécules Biologiques UMR6098 ( CNRS, U. de Provence, U. de la Méditerranée ) Case 932 163 Avenue de Luminy 13288 Marseille cedex 9 France Tel : +33(0) 491 82 55 93 Fax: +33(0) 491 26 67 20 e-mail: miguel.ortiz-lombar...@afmb.univ-mrs.fr Web: http://www.pangea.org/mol/spip.php?rubrique2 -- This message has been scanned for viruses and dangerous content by MailScanner, and is believed to be clean.
Re: [ccp4bb] PHENIX BSS refinement
Hi Miguel, there are limits and they are applied to total B-factor: - minimal Biso should be positive; - anisotropic ADP matrix is always enforced to be positive definite. Pavel. Thanks, it is clear now. I have a last concern, though. Does phenix impose minimal values to the refined ADPs? I guess it does not, for I have found nothing like that in the documentation. But, if it does so, are they imposed to the individual B-factors or the B-overall is also taken into account? In the former case you might have problems with the refinement in a case as the one I described in my previous e-mail (coming from a model with artificially low, 'residual' B-factors).
[ccp4bb] PHENIX BSS refinement
We have seen that the recent versions of phenix (1.3) the structure refinement ends with a bulk solvent correction step, something I found quite unusual (absent in other refinement programs and in older versions of phenix, 1.24.1). We have also seen that the B factor jumps up in this last step (see bellow). Is anybody aware of this? Do you know how to get rid of this final bss step? This step is redundant, as bss is also done at the beginning of each ref cycle. Header of a pdb file after refinement with phenix version 1.3: REMARK R-factors, x-ray target values and norm of gradient of x-ray target REMARK stage r-work r-free xray_target_w xray_target_t REMARK 0 : 0.3729 0.3776 5.594260e+00 5.604698e+00 REMARK 1_bss: 0.3135 0.3236 5.460255e+00 5.490057e+00 BSS refinement step REMARK 1_xyz: 0.3096 0.3226 5.451384e+00 5.487128e+00 refinement of coordinates REMARK 1_adp: 0.2608 0.2858 5.300551e+00 5.386607e+00 refinement of ADPs REMARK 2_bss: 0.2601 0.2843 5.299065e+00 5.384776e+00 REMARK 2_xyz: 0.2545 0.2847 5.285488e+00 5.386100e+00 REMARK 2_adp: 0.2460 0.2821 5.261319e+00 5.375879e+00 REMARK 3_bss: 0.2456 0.2788 5.257944e+00 5.366116e+00 REMARK 3_xyz: 0.2457 0.2809 5.261113e+00 5.371113e+00 REMARK 3_adp: 0.2427 0.2804 5.252685e+00 5.370838e+00 REMARK 3_bss: 0.2427 0.2804 5.252685e+00 5.370838e+00 ??? Bfactors REMARK stage b_max b_min b_ave REMARK 0 : 30.00 30.00 30.00 REMARK 1_bss: 30.00 30.00 30.00 REMARK 1_xyz: 30.00 30.00 30.00 REMARK 1_adp: 118.26 1.18 35.04 REMARK 2_bss: 118.26 1.18 35.04 REMARK 2_xyz: 118.26 1.18 35.04 REMARK 2_adp: 157.72 0.00 37.99 REMARK 3_bss: 157.72 0.00 37.99 REMARK 3_xyz: 157.72 0.00 37.99 REMARK 3_adp: 166.60 0.00 38.73 REMARK 3_bss: 203.00 36.40 75.12 ?? Thanks a lot for your help. Jose M Casasnovas Centro Nacional de Biotecnología (lab. B16) CSIC, Campus UAM Darwin 3 28049 Madrid Spain Ph. 34 915854917(8) Fax. 34 915854506 email: jcasasno...@cnb.csic.es !! NEW E-MAIL
Re: [ccp4bb] PHENIX BSS refinement
Hi Jose, We have seen that the recent versions of phenix (1.3) the structure refinement ends with a bulk solvent correction step, something I found quite unusual (absent in other refinement programs and in older versions of phenix, 1.24.1). - the software is evolving and gets better over time -:) That is, version 1.3 is supposed to be better than 1.24.1, and the latest 1.4 should be better than 1.3. - it is good to update bulk solvent correction at the end of refinement. For example, the refinement of coordinates may invalidate the mask which in turn invalidates the solvent contribution Fbulk. We have also seen that the B factor jumps up in this last step (see bellow). Is anybody aware of this? This is because the trace of overall anisotropic scale matrix is added to atomic B-factors and subtracted from that matrix. This is exactly what CNS does (at least version 1.1). Do you know how to get rid of this final bss step? It is not a good idea, so phenix.refine does not have an option to turn it off. This step is redundant, as bss is also done at the beginning of each ref cycle. It is not redundant: see above. FYI: there is PHENIX bulletin board: http://www.phenix-online.org/mailman/listinfo/phenixbb Pavel.
