Hi Maddy, It could well be that there is an issue with the data. I was wondering if you'd seen Sebastian Morin's consistency testing work before? This is part of relax and can be used to check the data. See http://www.nmr-relax.com/refs.html#Morin09. Considering this is the R2, there could have been sample heating problems. Are these experiments temperature compensated? And did you calibrate your temperatures on MeOH using a single scan directly after a partial R2 experiment completed?
Regards, Edward On 17 June 2011 18:53, Maddy Strickland <[email protected]> wrote: > Hi Edward, > > So I've got wild type = oblate chosen, and mutant = ellipsoid chosen. > When you look at the wild type, this gets the strange low values in both > oblate and ellipsoid, but not prolate/sphere. The mutant has beautiful > values in all. I'm wondering whether the wild type data is just not as > good. I might need to go over it - I did it much earlier, so my peak > picking might be wrong. I've found that R1 and NOE values agree very > well, as they should, but R2 are very different - since these experiments > have the largest noise, it may be due to the way I peak picked them - with > the mutant I actually manually peak picked, whereas with the wild type I > didn't I just picked every peak on the same spot, using the peak in the > first delay as a reference. I think the latter technique may help with > the R2 values. Chemical exchange should be similar in both apart from the > mutated loop, where it disappears in the mutant. > > I think basically - the two proteins where I manually peak picked I got > the best results, and in the fastest time too, with the fewest rounds. > The two proteins where I didn't worry as much about the T2 data, I got > extremely long calculation times and I ended up having to delete some > residues' data in order to allow the programme to finish. I think it > might be best if I go back and look at the T2 data again and rerun for the > latter two proteins. > > Maddy > > ---------------------------- Original Message ---------------------------- > Subject: Re: [Fwd: Re: Extracting results - eg sphere] > From: "Edward d'Auvergne" <[email protected]> > Date: Thu, June 16, 2011 8:10 am > To: "Maddy Strickland" <[email protected]> > Cc: [email protected] > -------------------------------------------------------------------------- > > Hi, > > It is strange that the S2 values are so far off. Are they the same if > you look at both ellipsoids or both prolate spheroids? How do the > diffusion tensors compare? Did you use single-scan interleaving, > temperature compensation blocks, and per experiment MeOH temperature > calibration? Do you see extensive chemical exchange or nanosecond > motions? How did you use relax, and which version of relax are you > using? How do both sets of NOEs compare, is the average identical in > both systems? And are the R1 and R2 averages the same? > > For the global models that are not in the 'final' run, it is true that > these have no errors. The reason is that Monte Carlo simulations are > expensive, computationally wise, so they are only used at the very end > once all model selection is performed. You can use relax to calculate > the errors for each model anyway. If you go to the last round_*/opt > directory, you can use a custom script there. Try something like: > > # Load the program state. > state.load('state') > > # Monte Carlo simulations. > monte_carlo.setup(number=500) > monte_carlo.create_data() > monte_carlo.initial_values() > minimise('newton') > eliminate() > monte_carlo.error_analysis() > > # Save the program state. > state.save('state_errors', force=True) > > You will have to play with this, as you might have to fix the > diffusion tensor if calculations are far too long. Or there might be > other option you'll need to play with too, but hopefully together with > the relax manual that shouldn't be too hard to work out. > > Regards, > > Edward > > > > > On 15 June 2011 15:06, Maddy Strickland <[email protected]> wrote: >> Yes, I want to look at all the different models - sphere/ellipsoid/prolate >> etc. to see how S2 values differ. I've got two proteins - one of which is >> a mutant of another (only a few residues mutated), but they came out with >> different models, one with ellipsoid and one with prolate. When I >> compared the S2 values for the two, one was about 0.5 different on every >> value (the prolate model was completely off). When looking at the >> spherical model it is much closer to what is expected. (I have four >> homologous proteins and all are very similar except this prolate model, so >> I can't work out why this is so different and I can't work out why it has >> been chosen with such strange S2 values. I wanted to compare all four >> spherical models instead of using anisotropic ones. I have worked out how >> to do this, but unfortunately, I think S2 error and Rex error is >> calculated right at the end in the 'final' round as I can't seem to >> extract this from 'opt' folders in the final round of 'sphere' or >> 'prolate' for example. >> >> Any ideas? Is error calculated in each round, or is it calculated right >> at the end and therefore impossible to collect for individual >> 'sphere'/'prolate' etc. that haven't been chosen? >> >> Maddy >> >> ---------------------------- Original Message ---------------------------- >> Subject: Re: Extracting results - eg sphere >> From: "Edward d'Auvergne" <[email protected]> >> Date: Tue, June 14, 2011 10:17 am >> To: "Maddy Strickland" <[email protected]> >> "Michael Bieri" <[email protected]> >> Cc: [email protected] >> -------------------------------------------------------------------------- >> >> Hi Maddy, >> >> It should be possible, but you may have to manually modify your >> extraction scripts. All of the results files are saved, so all of the >> data is there. I'm assuming you'd like to compare the different >> diffusion tensor results, is that correct? You will need to choose >> which file you look at. Just look through the directories and pick >> the final round for each tensor, and go into the 'opt' directory. The >> results.bz2 file in that directory will be the one you want. Maybe >> Michael can help with the modification of his script, if you are >> having troubles. >> >> Regards, >> >> Edward >> >> >> >> On 9 June 2011 13:03, Maddy Strickland <[email protected]> wrote: >>> Hello everyone, >>> >>> I was wondering if there was a way to extract results, say just from the >>> spherical folder or just from the ellipsoid folder for example, displayed >>> as s2.txt, rex.txt etc. like with the final data extraction script, but >>> for a different model than is selected by the program? >>> >>> >>> Madeleine Strickland >>> >>> MCJC Group >>> N317, School of Chemistry, Bristol University >>> >>> >>> _______________________________________________ >>> relax (http://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 >>> >> >> >> Madeleine Strickland >> >> MCJC Group >> N317, School of Chemistry, Bristol University >> >> >> _______________________________________________ >> relax (http://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 >> > > > Madeleine Strickland > > MCJC Group > N317, School of Chemistry, Bristol University > > _______________________________________________ relax (http://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
