Hi, I would recommend reading Sebastien Morin's paper about the consistency testing to understand how to interpret the plots. He also gives the relevant references for the different parts of the tests where you will find more details. I would recommend it. As for the model elimination, this is completely explained in my 2006 paper on model-free model elimination (http://dx.doi.org/10.1007/s10858-006-9007-z). The relevant rule from that paper is that your internal correlation times are 1.5 greater than the global correlation time of the molecule. With today's limitations of what you can measure, you should not see motion on these time scales as they are hidden by experimental noise.
Looking at your results file, there are many more problems than just this model elimination. The chi2 value is over 2 million, which indicates a very, very bad fit! Now that I looked more carefully at the results file, I can see the exact problem. I would recommend multiplying your proton frequencies by 1e6, as they should be in Hz ;) I think I'll add some checks to relax for the input frequency. Regards, Edward On 5 June 2012 13:35, Martin Ballaschk <[email protected]> wrote: > Hi Edward, > > thank you for the reply. I will file a bug report soon. > > I'll answer below: > > On 05.06.2012, at 11:41, Edward d'Auvergne wrote: >> Do you have information about which version of relax you are using? > > It's 1.3.16 – newly downloaded from the main page. > >> [...] >> Note that this RelaxFault is not much of an issue. You can see the >> exact failure point by looking into the >> auto_analyses/dauvergne_protocol.py file. > > The "relax controller" also showed me the point of failure. I copied the last > 100.000 lines of messages into this file: > https://www.dropbox.com/sh/ijah1ll37o9xmsy/le6rGRguU5/mf_firstrun_log.txt > > That's the last lines: > > relax> pymol.macro_write(data_type='amp_fast', style='classic', > colour_start=None, colour_end=None, colour_list=None, file=None, > dir='/faust/TENSOR/sh3_EdA/final/pymol', force=True) > Exception raised in thread. > > Traceback (most recent call last): > File "/usr/local/relax/gui/analyses/execute.py", line 88, in run > self.run_analysis() > File "/usr/local/relax/gui/analyses/auto_model_free.py", line 765, in > run_analysis > dauvergne_protocol.dAuvergne_protocol(pipe_name=self.data.pipe_name, > results_dir=self.data.save_dir, diff_model=self.data.global_models, > mf_models=self.data.mf_models, local_tm_models=self.data.local_tm_models, > grid_inc=self.data.inc, diff_tensor_grid_inc=self.data.diff_tensor_grid_inc, > mc_sim_num=self.data.mc_sim_num, max_iter=self.data.max_iter, > conv_loop=self.data.conv_loop) > File "/usr/local/relax/auto_analyses/dauvergne_protocol.py", line 223, in > __init__ > self.execute() > File "/usr/local/relax/auto_analyses/dauvergne_protocol.py", line 710, in > execute > self.write_results() > File "/usr/local/relax/auto_analyses/dauvergne_protocol.py", line 858, in > write_results > self.interpreter.pymol.macro_write(data_type='amp_fast', dir=dir, > force=True) > File "/usr/local/relax/prompt/pymol_control.py", line 266, in macro_write > pymol_control.macro_write(data_type=data_type, style=style, > colour_start=colour_start, colour_end=colour_end, colour_list=colour_list, > file=file, dir=dir, force=force) > File "/usr/local/relax/generic_fns/pymol_control.py", line 458, in > macro_write > commands = create_macro(data_type=data_type, style=style, > colour_start=colour_start, colour_end=colour_end, colour_list=colour_list) > File "/usr/local/relax/generic_fns/pymol_control.py", line 352, in > create_macro > commands = macro(data_type, style, colour_start, colour_end, colour_list) > File "/usr/local/relax/specific_fns/model_free/macro_base.py", line 551, in > create_macro > self.classic_style(data_type, colour_start, colour_end, colour_list, > spin_id) > File "/usr/local/relax/specific_fns/model_free/macro_base.py", line 242, in > classic_style > raise RelaxFault > > Storing the relax state in the file 'relax_state_20120602_135951.bz2'. > > RelaxFault: RelaxError: Impossible to be here, please re-run relax with the > '--debug' flag and summit a bug report at https://gna.org/projects/relax/. > > >> [...] This means that you have the results for >> the completed analysis sitting on your computer! > > Yes, you're right. I can see the the final result files in the respective > directories ("./final"). Unfortunately, something is still wrong, and I > suspect it's my data. > > relax obviously went for the ellipsoid model, but during the runs, most of > the spin systems got eliminated. That means, for most of the parameters I > don't have any values (see > https://www.dropbox.com/sh/ijah1ll37o9xmsy/UgQuNzBSwl/final ). > > I don't really understand what exactly is tested against what when relax > tells me e.g.: > > Data pipe 'm5': The ts value of 8.5026e-10 is greater than 6.377e-10, > eliminating spin system '#1aey-NMR_mol1:30&:ILE@410&@N'. > > I'm sure that's described somewhere in the manual or the thesis, but I > couldn't find it. > > *** > > So what can I do about this. The obviuos thing is to validate my data (again) > and finally do the temperature calibration at the 750 machine. Isn't it odd > that the R1, R2 methanol calibrations we did showed no differences in ppm, > regardless of the duration of the R2 measurements? > > Is maybe the peak maximum searching in CCPN introducing systematic error by > picking positive noise on top of the peaks? > > The NOE error calculation in CCPN is somewhat opaque and is based on getting > the spectrum noise. Could a under / overestimation of the NOE errors lead to > such a massive elimination of spin models? > > I don't really know where to start looking for errors. > > Cheers > Martin > > > > > >> On 5 June 2012 11:00, Martin Ballaschk <[email protected]> wrote: >>> >>> Hello Edward, hello relax-users, >>> >>> for five days relax calculated mf models for SH3 and then something went >>> wrong. It stopped with >>> >>>> RelaxFault: RelaxError: Impossible to be here, please re-run relax with >>>> the '--debug' flag and summit a bug report at >>>> https://gna.org/projects/relax/ >>> >>> A look at the relax messages tells me that almost all spin got eleminated. >>> The "debugging run" is still running. >>> >>> I will try to give a detailed description of what we've done to collect the >>> input data and maybe you can give me a hint what may be a problem; if it's >>> something with the sample, the measuring setup or the processing workflow. >>> >>> ==== >>> >>> 1.) Protein / sample >>> >>> * SH3 from chicken alpha-spectrin, 62 residues (1AEY.pdb) >>> * 2H/13C/15N and metyhl-group labeled >>> * concentration: ~ .5 mM >>> * our "all-purpose model protein" for testing new stuff >>> >>> >>> 2.) Data collection >>> >>> * two fields: 600 MHz & 750 MHz (Bruker spectrometers with cryo probes) >>> * 600 machine: >>> http://www.fmp-berlin.de/schmieder/nmr_facility/av600sel.htm >>> * 750 machine: >>> http://www.fmp-berlin.de/schmieder/nmr_facility/av750.htm >>> * temperature: 300.1 K >>> o comparison between between T1 and T2 temperatures /w methanol >>> gave *no* differences in temperature at the 600 MHz machine (!) >>> o 750 MHz spectrometer needs still to be tested though >>> o temperature differences in HetNOE also need to be tested >>> * single-scan interleaved recording was performed for T1, T2 and HetNOE >>> measurements >>> * T1: delays between 12 and 1500 ms, 8 data points, 4 of them replicated >>> * T2: delays between 17 and 217 ms, 8 data points, 4 of them replicated >>> >>> >>> 3.) Data processing >>> >>> * processing of all spectra: Bruker TOPSPIN 3.1 >>> * peak picking with CCPN 2.2.1: >>> o T1, T2 (pseudo-3D) picked with „Follow intensity changes“ function >>> o HetNOE (2D planes) picked with „Heteronuclear NOE“ function >>> * T1, T2 analysed with relax' fitting function >>> * consisteny testing performed with Sebastien Morins script >>> (although I'm not sure about the interpretation of the results, >>> also: the molecule is very small and tau_m < 5ns) >>> >>> >>> 4.) Model-free analysis within relax GUI >>> >>> * imported all relaxation data as described >>> * CSA, bond length values: default values taken >>> * unit vectors from 1AEY.pdb (averaged in relax) >>> >>> >>> All the relax data of this run including the results files and consistency >>> testing plots have been made publicly available: >>> https://www.dropbox.com/sh/ijah1ll37o9xmsy/OybJrGT13e >>> >>> >>> I would be very grateful if you could have a look at the data and tell me: >>> is it me or is it a bug? >>> >>> Cheers >>> Martin > _______________________________________________ relax (http://www.nmr-relax.com) This is the relax-users mailing list [email protected] To unsubscribe from this list, get a password reminder, or change your subscription options, visit the list information page at https://mail.gna.org/listinfo/relax-users
