- phenixbb - phenix-online.org

Re: [phenixbb] The performance of phenix.refine
by Ralf W. Grosse-Kunstleve 22 Mar '07

22 Mar '07

> Did anybody actually compare the performance among phenix.refine, CNS > and refmac5? I am trying the phenix.refine from the latest CCI > bundle for my refinement. I have a relative large structure (800 kD > unique) and unit cell (140, 160, 550, 90, 90, 90). The data is at > low resolution ( 3.8 A) with experimental phases. My impression is > that phenix.refine takes a much longer time to finish the jobs, say > rigid body refinement or SA refinement, and it also takes a lot more > memories ( in my case, phenix.refine uses more than 700 MB, but other > programs use less than 200 MB). We have been concentrating on features and a high level of automation much more than raw speed and memory efficiency. Given that memory is so cheap these days (ca. $80/GB) there isn't much incentive for us to trade developer time for a smaller memory footprint. Speed optimizations are a bit higher up on the list. > Is this because the python code in > phenix.refine? CNS and refmac5 are written in FROTRAN, I think. Just > curious. phenix.refine doesn't spend much time in the Python code. It is a long while ago I did careful timings, but I'd guess it is significantly less than 10% of the time. C++ is intrinsically almost as fast as FORTRAN if the same development time is invested in the code. I.e. what you see is not a language issue, just a tradeoff features vs. optimization. We'll work on making phenix.refine a bit faster in the future, but at the minute more features are still our focus. Ralf

1 0

The performance of phenix.refine
by Jianghai Zhu 22 Mar '07

22 Mar '07

Hi, Did anybody actually compare the performance among phenix.refine, CNS and refmac5? I am trying the phenix.refine from the latest CCI bundle for my refinement. I have a relative large structure (800 kD unique) and unit cell (140, 160, 550, 90, 90, 90). The data is at low resolution ( 3.8 A) with experimental phases. My impression is that phenix.refine takes a much longer time to finish the jobs, say rigid body refinement or SA refinement, and it also takes a lot more memories ( in my case, phenix.refine uses more than 700 MB, but other programs use less than 200 MB). Is this because the python code in phenix.refine? CNS and refmac5 are written in FROTRAN, I think. Just curious. Jianghai +++++++++++++++++++++++++++++++ Jianghai Zhu, Ph.D CBR Institute for Biomedical Research Department of Pathology Harvard Medical School 200 Longwood Ave., Boston, MA 02115 Ph: 617-278-3211 Fx: 618-278-3232 +++++++++++++++++++++++++++++++

1 0

Re: [phenixbb] freeR syntax question in phenix.refine
by Ralf W. Grosse-Kunstleve 20 Mar '07

20 Mar '07

> I was running refinement from an mtz file from sigmaa in ccp4 with an > appropriate FreeR flag designated in the MTZ file. phenix.refine > worked fine and seemed to exclude the test reflections, but I wanted > to know what these commands do. Don't! :-) > r_free_flags { > file_name = "S19_sigmaa1.mtz" > label = "FreeR_flag" > test_flag_value = 0 --> don't know what this means? > disable_suitability_test = False --> implications of this? phenix.refine is designed to figure everything out automatically. You will have to specify file_name and label only if you have multiple R-free-flag arrays in the reflection files you give to phenix.refine, which is very unusual. phenix.refine does some sanity checks to determine if arrays that "look like" R-free-flag arrays are actually suitable for refinement, and reject them if "it doesn't think so". It will then tell you that you can override the sanity checks by setting disable_suitability_test to True. Again, this is very unusual. The test_flag_value could potentially be important. phenix.refine analyses the putative R-free-flag array(s) to guess if they came from CNS, where 0 usually means work set, 1 free set, or CCP4, which mostly uses values from 0 to 19, 0 is the default free set in REFMAC. Therefore, roughly, if phenix.refine "thinks" it is a CNS R-free-flag array it assigns test_flag_value=1, or 0 if it "thinks" it is a CCP4 array. Only if you used REFMAC before for refinement, and only if you manually instructed REFMAC to use a value different from 0 for the free set, will you have to manually tell phenix.refine about it, e.g. via r_free_flags.test_flag_value=19. If you are interested in more details, inspect the script iotbx/iotbx/reflection_file_utils.py, which does all the guesswork. Look for "get_r_free_flags_scores".

1 0

freeR syntax question in phenix.refine
by Simon Newstead 20 Mar '07

20 Mar '07

Dear BB, Sorry for the obvious question, but I couldn't find a description of the syntax for the FreeR flag function in phenix.refine. Is there a paragraph on the phenix web site that explains what I should put in these fields: r_free_flags { file_name = "S19_sigmaa1.mtz" label = "FreeR_flag" test_flag_value = 0 --> don't know what this means? disable_suitability_test = False --> implications of this? I was running refinement from an mtz file from sigmaa in ccp4 with an appropriate FreeR flag designated in the MTZ file. phenix.refine worked fine and seemed to exclude the test reflections, but I wanted to know what these commands do. Cheers, Simon.

2 1

Twinned Refinement
by Axelrod, Herbert 19 Mar '07

19 Mar '07

Hi, I was wondering if anyone has tested the new version of phenix.refine for the refinement of twinned diffraction data. Last week at Asilomar, Pavel presented a nice talk at Asilomar on the capabilities of phenix.refine to handle twinned data. Best wishes, Herbert L. Axelrod Stanford Synchrotron Radiation Lab

2 3

phenix.refine crash
by Jianghai Zhu 19 Mar '07

19 Mar '07

Hi, I was using phenix.refine from the latest cci apps bundle to do tls +individual_sites+individual_adp refinement. The program crashed with the following error message. ----------Target weights (before refinement)---------- Traceback (most recent call last): File "/nfs/home/jzhu/cci_apps_sources/phenix/phenix/command_line/ refine.py", line 5, in <module> command_line.run(command_name="phenix.refine", args=sys.argv[1:]) File "/nfs/home/jzhu/cci_apps_sources/phenix/phenix/refinement/ command_line.py", line 75, in run call_back_after_monitor_collect=call_back_after_monitor_collect) File "/nfs/home/jzhu/cci_apps_sources/phenix/phenix/refinement/ driver.py", line 1108, in run log = log) File "/nfs/home/jzhu/cci_apps_sources/phenix/phenix/refinement/ strategies.py", line 245, in refinement_machine log = log) File "/nfs/home/jzhu/cci_apps_sources/phenix/phenix/refinement/ weight_xray_chem.py", line 222, in __init__ compute_gradients = True) File "/nfs/home/jzhu/cci_apps_sources/mmtbx/mmtbx/restraints.py", line 144, in energies_adp_aniso xray_structure = xray_structure, File "/nfs/home/jzhu/cci_apps_sources/cctbx/cctbx/adp_restraints/ __init__.py", line 109, in __init__ self.check_flags(fl_i) File "/nfs/home/jzhu/cci_apps_sources/cctbx/cctbx/adp_restraints/ __init__.py", line 177, in check_flags assert not fl.use_u_aniso() AssertionError Anyone has any idea what is happening? Thanks. Jianghai +++++++++++++++++++++++++++++++ Jianghai Zhu, Ph.D CBR Institute for Biomedical Research Department of Pathology Harvard Medical School 200 Longwood Ave., Boston, MA 02115 Ph: 617-278-3211 Fx: 618-278-3232 +++++++++++++++++++++++++++++++

5 9

More stringent water picking
by Schubert, Carsten [PRDUS] 19 Mar '07

19 Mar '07

Hi, is it possible to increase the stringency criteria for picked waters, besides increasing the sigma level? E.G. something on the order of the H-bonding criterion found in CNS? I find that the automated water-picking routine gives me quite a few waters, which are not necessarily justified by visual inspection of the maps. Thanks Carsten

5 4

Re: [phenixbb] More stringent water picking
by Schubert, Carsten [PRDUS] 19 Mar '07

19 Mar '07

Thanks Guys, would it be possible to implement the H-bonding criteria? From my experience with CNS it worked quite nicely. Right now I end up with ~350 waters in a 2.3 A structure of a 30kDa protein, which is a bit much. -----Original Message----- From: phenixbb-bounces(a)phenix-online.org [mailto:[email protected]] Sent: Monday, March 19, 2007 1:02 PM To: PHENIX user mailing list Subject: Re: [phenixbb] More stringent water picking Hi Carsten, thanks for your question! The main water picking criteria in phenix.refine are: 1) Fo-Fc map; 2) 2Fo-Fc map; 3) macromolecule - water distance; 4) water-water distance; 5) B-factors and occupancies max/min values; plus some more of expert level ... There is no specific H-bonding criteria implemented. You can play with sigma cutoff for both maps and with the distances above. In most of the cases it the reasonable behavior. I would be interested in Paul's (Adams) comments about his experience of usefulness of H-bonding criteria in CNS. Thanks! Pavel. Schubert, Carsten [PRDUS] wrote: Hi, is it possible to increase the stringency criteria for picked waters, besides increasing the sigma level? E.G. something on the order of the H-bonding criterion found in CNS? I find that the automated water-picking routine gives me quite a few waters, which are not necessarily justified by visual inspection of the maps. Thanks Carsten _____ _______________________________________________ phenixbb mailing list phenixbb(a)phenix-online.org <mailto:[email protected]> http://www.phenix-online.org/mailman/listinfo/phenixbb <http://www.phenix-online.org/mailman/listinfo/phenixbb>

1 0

Re: [phenixbb] OS X installation problem
by Ralf W. Grosse-Kunstleve 17 Mar '07

17 Mar '07

> powerpc-apple-darwin8-g++-4.0.0 (GCC) 4.0.0 20041026 (Apple Computer, > Inc. build 4061) There is a more recent Xcode release that works here. I.e. if you still want to compile from sources, upgrade the Xcode tools. You can get them for free at Apple's developer web site. However... > is that the 10.3.9 bundle? Yes, exactly. I should also work under 10.4. I recommend trying this first. It is also most faster than a source installation (just a few seconds). Ralf

1 0

Re: [phenixbb] OS X installation problem
by Ralf W. Grosse-Kunstleve 17 Mar '07

17 Mar '07

I have not seen this error before. Could you send me the output of c++ --version Note that we have binary bundles for PowerPC Macs. Did you try that one? Ralf

2 1