- phenixbb - phenix-online.org

PDF slides: introduction to refinement, phenix.refine and why automation is important (PHENIX)
by Pavel Afonine 09 Mar '09

09 Mar '09

Dear PHENIX community: I would like to draw your attention to these two PDF files with the slides that I recently made for a PHENIX workshop (workshop on Neutron crystallography, March 5-6, Japan). The first file (69 of very dense slides) makes a general introduction to crystallographic structure refinement (certainly biased towards how things are done in phenix.refine), plus some introduction to phenix.refine functionality. The second PDF file talks about PHENIX and "quick validation" with the strong accent of why automation is important. In general, I don't like and normally don't do such advertisements, but in this case I believe I can make an exception since I spent too much "of my software development" time on making these slides and they might be instructive for (new) PHENIX users. The links are (the first two!): http://www.phenix-online.org/presentations/neutron_japan_2009/ Thanks, Pavel. PS> Two slides in the first file are borrowed from Sasha Urzhumtsev, and 2-3 slides are borrowed from Paul Adams (second file).

2 2

Re: [phenixbb] phenixbb Digest, Vol 40, Issue 7
by Pascal Egea 08 Mar '09

08 Mar '09

Pavel and Paul, this is really awesome.Thanks a lot. Pascal Egea UCSF Department of Biochemistry and Biophysics Laboratory of Robert Stroud On Sun, Mar 8, 2009 at 1:00 PM, <phenixbb-request(a)phenix-online.org> wrote: > Send phenixbb mailing list submissions to > phenixbb(a)phenix-online.org > > To subscribe or unsubscribe via the World Wide Web, visit > http://www.phenix-online.org/mailman/listinfo/phenixbb > or, via email, send a message with subject or body 'help' to > phenixbb-request(a)phenix-online.org > > You can reach the person managing the list at > phenixbb-owner(a)phenix-online.org > > When replying, please edit your Subject line so it is more specific > than "Re: Contents of phenixbb digest..." > > > Today's Topics: > > 1. PDF slides: introduction to refinement, phenix.refine and why > automation is important (PHENIX) (Pavel Afonine) > > > ---------------------------------------------------------------------- > > Message: 1 > Date: Sat, 07 Mar 2009 23:41:35 -0800 > From: Pavel Afonine <PAfonine(a)lbl.gov> > Subject: [phenixbb] PDF slides: introduction to refinement, > phenix.refine and why automation is important (PHENIX) > To: phenixbb(a)phenix-online.org > Message-ID: <49B376AF.1020803(a)lbl.gov> > Content-Type: text/plain; charset=ISO-8859-1; format=flowed > > Dear PHENIX community: > > I would like to draw your attention to these two PDF files with the > slides that I recently made for a PHENIX workshop (workshop on Neutron > crystallography, March 5-6, Japan). > > The first file (69 of very dense slides) makes a general introduction to > crystallographic structure refinement (certainly biased towards how > things are done in phenix.refine), plus some introduction to > phenix.refine functionality. > > The second PDF file talks about PHENIX and "quick validation" with the > strong accent of why automation is important. > > In general, I don't like and normally don't do such advertisements, but > in this case I believe I can make an exception since I spent too much > "of my software development" time on making these slides and they might > be instructive for (new) PHENIX users. > > The links are (the first two!): > > http://www.phenix-online.org/presentations/neutron_japan_2009/ > > Thanks, > Pavel. > > PS> Two slides in the first file are borrowed from Sasha Urzhumtsev, and > 2-3 slides are borrowed from Paul Adams (second file). > > > > > ------------------------------ > > _______________________________________________ > phenixbb mailing list > phenixbb(a)phenix-online.org > http://www.phenix-online.org/mailman/listinfo/phenixbb > > > End of phenixbb Digest, Vol 40, Issue 7 > *************************************** >

1 0

Re: [phenixbb] refinement using a multiprocessor machine or on a cluster
by Pavel Afonine 07 Mar '09

07 Mar '09

Hi Gino, phenix.refine does not use any advantage of multiple available CPUs. Sorry. Pavel. On 3/4/09 11:26 AM, Gino Cingolani wrote: > Dear Phenix community, > > I was wondering if it is possible to run a refinement job on multiple > processor. The refinement of my 13,000 asymmetric unit takes about 2 > days. It could be greatly speeded up by taking advantage of the 4 CPUs > in my PC. The phenix manual suggest to use > nproc=3 > to run using a multiprocessor machine or on a cluster. > But this applies only to LigandFit and AutoBuild Wizards. > What about simple refinement? > > Thanks in advance, > > Gino > >

2 1

fwd: TLS without isotropic adp refinement
by Nathaniel Echols 07 Mar '09

07 Mar '09

(I think the server must have a lag time between adding a new email address and accepting posts from it. . .) > From: Roeland Boer <rbocri(a)ibmb.csic.es> > Date: March 5, 2009 5:29:19 AM PST > To: phenixbb(a)phenix-online.org > Subject: TLS without isotropic adp refinement > > Dear all, > > I'm trying to refine a structure using TLS on some of the domains > without refining the isotropic B factors of the atoms of the full > structure. I haven't been able to do this: the isotropic B factors > seem to be refined per residue when I include TLS. If I don't use > TLS, the B factors are not refined. I've looked around in the > archives and browsed the manual without luck. Any suggestions would > be highly appreciated. I include the parameter file I've used. Quoting Pavel (since he's teaching a workshop): "phenix.refine always writes out the total B-factors and not split them. The total B-factor in phenix.refine is defined as: Utotal = Utls + Ulocal + Ucryst." So the atomic B-factors in your structure are not being refined individually, but they change anyway because the contribution of the TLS refinement is included. -Nat > > Kind regards, > Roeland Boer. > > > ++++++++++++++ Parameters ++++++++++++++++ > > refinement.main.number_of_macro_cycles=4 > refinement.main.ncs = True > refinement.input.xray_data { > labels=F > low_resolution=20.0 > high_resolution=3.6 > } > refinement.refine { > strategy=tls+individual_sites > adp { > tls="chain A and (resid 4:132 or resid 301)" > tls="chain B and (resid 4:132 or resid 301)" > tls="chain C and (resid 4:132 or resid 301)" > tls="chain X and (resid 4:132 or resid 301)" > tls="chain Y and (resid 4:132 or resid 301)" > tls="chain Z and (resid 4:132 or resid 301)" > } > } > refinement.ncs { > find_automatically = False > restraint_group { > reference = chain A and resid 133:203 > selection = chain B and resid 133:203 > selection = chain C and resid 133:203 > selection = chain X and resid 133:203 > selection = chain Y and resid 133:203 > selection = chain Z and resid 133:203 > } > restraint_group { > reference = chain B and (resid 4:132 or resid 301) > selection = chain A and (resid 4:132 or resid 301) > selection = chain C and (resid 4:132 or resid 301) > selection = chain X and (resid 4:132 or resid 301) > selection = chain Y and (resid 4:132 or resid 301) > selection = chain Z and (resid 4:132 or resid 301) > } > } > refinement.output.prefix=3.6-tlsnter-indsite-inputb-autowght-ncs > -------------------- Parameters -------------------------- > > ------------------- Nathaniel Echols Lawrence Berkeley Lab 510-486-5136 NEchols(a)lbl.gov

2 1

Re: [phenixbb] fwd: kicked maps
by Pavel Afonine 07 Mar '09

07 Mar '09

Hi Maia, I think it's easier to try out and have an answer right away rather than speculate. Why don't you just try and see if it helps? I can't tell in advance if it will help or not. Again, a "average kick map" (AKM) is computed as following: we create a large ensemble of structures (~500 structures) that are all randomly shaken. The shake amount (rmsd distortion introduced to coordinates) varies from 0 to 1.0A. Then for each structure the map is computed (mFo-DFc or 2mFo-DFc or any other requested by the user). And finally, all maps are averaged out to produce one "averaged kick map". Obviously, just from general considerations, such map is supposed to have less or no bias, less noise and enhanced signal, and potentially can clear up some initially bad densities. I would also try an "omit AKM", where you remove some model parts in questions and compute mFo-DF and 2mFo-DFc AK maps. Please use the latests CCI Apps for this since we recently fixed a number of problems with this. Pavel. >> Hi Pavel, I have a problem with long linkers between domains in a >> protein structure that is a biological hexamer. >> I don't have much density for the linkers. Should "kicked maps" help >> me to see the density? The resolution of the X-ray data is 2.3A. >> >> Maia >

1 0

default bondlengths - metal
by Engin Ozkan 06 Mar '09

06 Mar '09

Hi, all, Recently following the phenixbb, I was made aware of phenix.metal_coordination. So I just gave it a try for a structure of mine with sodium in it. Specifically, I decided to turn the use-default-bondlengths option on. The result was surprising, because the output file thinks that sodum to oxygen distances are supposed to be 1.9 Angstroms. Work done by Marjorie Harding and others have shown that this distance should be about 2.4 (Acta Cryst. D62 (2006), 678-682; for a summary of their numbers, http://tanna.bch.ed.ac.uk/newtargs_06.html). I wonder why phenix.metal_coordination chose 1.9 A as the default Na-O distance, or is the default option a different type of bond distance, I am not sure. Any clarification from the authors would be appreciated. Below is part of the output, showing the distance: refinement.geometry_restraints.edits { bond { action = *add atom_selection_1 = name NA and chain N and resname NA and resseq 1 atom_selection_2 = name O and chain A and resname PHE and resseq 76 distance_ideal = 1.920420 sigma = 0.01 } } Engin

5 7

fwd: phenix.autobuild- multiple models.
by Nathaniel Echols 04 Mar '09

04 Mar '09

> (The server bounced this message for some reason.) > From: Michael Karl Uhl <michael.uhl(a)uni-graz.at> > Date: March 3, 2009 11:55:16 PM PST > To: phenixbb(a)phenix-online.org > Subject: phenix.autobuild- multiple models > Reply-To: michael.uhl(a)a-b.at > > > Hello, > > I have not registered for this list until now, but have a very > urgent question concerning phenix.autobuild and multiple models as > input. Is there a possibility to merge models generated by different > phenix.autobuild runs. I have several overall_best-structure and > want to merge them into one single overall_best structure. I have > found the keyword "merge_models= True" but it does not really work > (Sorry, unknown file or keyword: merge_models=True). How should I > make the input in this case? > > > Thanks a lot, > > > Michael Uhl. > > ------------------- Nathaniel Echols Lawrence Berkeley Lab 510-486-5136 NEchols(a)lbl.gov

2 1

Re: [phenixbb] Please help with neutron composite SA-omit
by Pavel Afonine 03 Mar '09

03 Mar '09

Hi Anna, a "kick map" is computed as following: we create a large ensemble of structures (~500 structures) that are all randomly shaken. The shake amount (rmsd distortion introduced to coordinates) varies from 0 to 1.0A. Then for each structure the map is computed (mFo-DFc or 2mFo-DFc or any other requested by the user). And finally, all maps are averaged out to produce one "averaged kick map". Obviously, such map is supposed to have less or no bias, less noise, and potentially can clear up some initially bad densities. The is currently no paper published about this method; we are preparing the manuscript. There are scattered references and uses of AK maps, mostly by Dusan Turk (he originally implemented it in his program MAIN). Documentation is not yet updated, this is true. Sorry. In phenix.refine GUI you can just check the corresponding box to compute a kick map (under map menu). In command line version you need to supply a parameter file that will define such map: refinement.electron_density_maps { map { mtz_label_amplitudes = 2FOFCWT mtz_label_phases = PH2FOFCWT likelihood_weighted = True obs_factor = 2 calc_factor = 1 kicked = True fill_missing_f_obs_with_weighted_f_model = True } map { mtz_label_amplitudes = FOFCWT mtz_label_phases = PHFOFCWT likelihood_weighted = True obs_factor = 1 calc_factor = 1 kicked = True fill_missing_f_obs_with_weighted_f_model = True } } I was recently trying this with neutron maps and the maps are significantly better (computed with AK approach). Please use the latest version of CCI Apps for this. Let me know if you have any questions or problems. Pavel. On 2/24/09 7:58 AM, Anna Gardberg wrote: > Hi, Pavel. > > I'm not trying to bring up features, exactly, but rather trying to get > an unbiased, quantitative measure of the the deuterium atoms' nuclear > density. The idea is to compare the visibility of hydrogen or > deuterium atoms with x-ray and neutron diffraction, correlating this > parameter (perhaps in the form of OMIT density) to others. > > I have only briefly toyed with kick-maps, but the lack of > documentation (http://phenix-online.org/documentation/refinement.htm > doesn't say much about how best to use them) made it frustrating. Can > you recommend an article about kick maps? Or some suggestions on how > to best use this feature, please? > > > -Anna > > On Mon, Feb 23, 2009 at 11:46 PM, Pavel Afonine <PAfonine(a)lbl.gov> wrote: > >> Hi Anna, >> >> you can ask phenix.refine produce "kick maps" that are supposed to be less >> biased, contain less noise and, if you are lucky, might bring some more >> features (that were buried in noise before). >> >> Let me know if you have any questions. >> >> Pavel. >> >> >> On 2/23/09 11:40 AM, Anna Gardberg wrote: >> >>> Hello. >>> >>> I am attempting to create an SA-omit composite map with neutron >>> scatterers. >>> >>> Unfortunately, >>> >>> phenix.autobuild data= "neutron.mtz" model="neutron_004.pdb" >>> composite_omit_type=sa_omit scattering_table=neutron >>> >>> results in an error: >>> >>> "Sorry, unknown file or keyword: scattering_table=neutron >>> Possibilities... (use 'phenix.autobuild --help all' to get full help): >>> general.resolve_pattern_command_list=neutron >>> density_modification.use_resolve_pattern=neutron >>> decision_making.semi_acceptable_r=neutron >>> decision_making.acceptable_r=neutron >>> multiple_models.multiple_models_starting_resolution=neutron >>> " >>> >>> Can someone here help, please? >>> >>> (For what it's worth: I'd use nCNS, but while CNS 1.21 (precompiled >>> binaries) works on my computer, CNS 1.1 version seems to compile, it >>> will not run. When I apply the nCNS patch to CNS 1.1, again it seems >>> to compile, but will not run. ) >>> >>> Regards, >>> Anna Gardberg >>> _______________________________________________ >>> phenixbb mailing list >>> phenixbb(a)phenix-online.org >>> http://www.phenix-online.org/mailman/listinfo/phenixbb >>> >>>

2 2

AutoBuild after resolve
by Mo Wong 28 Feb '09

28 Feb '09

Hi all, Can someone tell me where in the AutoBuild GUI I can set "phases_from_resolve"? I've looked for it in the "set variable" section, but it's not listed. Thanks!

2 3

rfree flag fractions
by Yusuf Akhter 28 Feb '09

28 Feb '09

Hi Everybody! I am using phenix version 1.3b for refining a structure at 3 A resolution and total number of total reflections are 12680. I have seen a provision in phenix to change the default maximal number of test flags to be generated and the fraction with this command line xray_data.r_free_flags.fraction=0.05 xray_data.r_free_flags.max_free=500 My queries are 1)when should somebody use this utility ?? 2)what is minimum number of Rfree flags one should use on this resolution and this total number of reflections?? Any kind of suggestions will be appreciated. Cheers! Yusuf

2 1