Hi Atul,
Troels pretty much hit the nail on the head with that response! I
have a few additional points below that I'll spread across a few
messages for better threads at
http://thread.gmane.org/gmane.science.nmr.relax.user/1718. The first
point is about the incredibly important error estimates:
On 21 August 2014 12:00, Troels Emtekær Linnet <tlinnet@xxxxxxxxxxxxx> wrote:
Dear Atul.
Thank you for a very descriptive problem.
I am sure this post will help others.
I will write an extended detailed description, in hope it can help others.
Most of it will be trivial for you.
Let us arm ourself with some reference information, to let the details
become clearer.
Since your last post, I have quickly expanded the wiki tutorial on
R1rho off resonance:
In the section: 1_setup_r1rho.py
http://wiki.nmr-relax.com/Tutorial_for_Relaxation_dispersion_analysis_r1rho_fixed_time_recorded_on_varian_as_sequential_spectra
And also the sample script you mention:
relax/sample_scripts/relax_disp/R1rho_analysis.py
http://svn.gna.org/viewcvs/*checkout*/relax/trunk/sample_scripts/relax_disp/R1rho_analysis.py?revision=HEAD
And we can use Figure 1 and 10 in the reference.
Palmer, A.G. & Massi, F. (2006). Characterization of the dynamics of
biomacromolecules using rotating-frame spin relaxation NMR
spectroscopy. Chem. Rev. 106, 1700-1719
DOI: http://dx.doi.org/10.1021/cr04042875
I have tried to reproduce that figure here:
http://wiki.nmr-relax.com/images/1/12/Fig1_Palmer_Massi_2006.png
------
Let us go through lines in the Sample script.
# Load the sequence.
sequence.read('fake_sequence.in', dir=DATA_PATH, res_num_col=1,
res_name_col=2)
-> This basically creates the spin-data structure, to store subsequent
information.
# Set the isotope information.
spin.isotope(isotope='15N')
-> By defining which isotope, we can convert to ppm, when we have the
spectrometer field.
-> The physical constants are pulled from: cat lib/physical_constants.py
Then comes a section, to load data.
In the sample script, the data is put in manually.
In the wiki tutorial, it is pulled from files, which are prepared in bash.
That is a 'preference' issue. The data I have analysed, were with 70+
spectrum.
So I preferred to do it in files.
Then we loop over the spectrums:
spectrum.read_intensities(file=file, dir=DATA_PATH, spectrum_id=id,
int_method='height')
spectrum.baseplane_rmsd(spectrum_id=id, error=rmsd)
-> The intensities, and its associated error. One can either measure
the RMSD of the spectrum,
or replicate spectrums. For R1rho, it will normally measuring the RMSD.
-> A quick way in linux is (for NMRPipe data):
showApod test.ft2 | grep "REMARK Automated Noise Std Dev in Processed
Data:" | awk '{print $9}'
Or in SPARKY:
http://wiki.nmr-relax.com/Tutorial_for_Relaxation_dispersion_analysis_cpmg_fixed_time_recorded_on_varian_as_fid_interleaved#Measure_the_backgorund_noise_.22RMSD.22_in_each_of_the_.ft2_files
For the RMSD of the baseplane noise estimate, it is much better to use
the Sparky feature where you measure the noise in small boxes. This
is why I discussed with Tom Goddard, in private, and helped him design
the 'rm' extension he developed and is now present in Sparky:
"""
Date: Fri, 24 Aug 2001 09:28:13 -0700 (PDT)
Message-Id: <200108241628.JAA300086@xxxxxxxxxxxxxxxxxxxxx>
X-Authentication-Warning: socrates.cgl.ucsf.edu: goddard set sender to
goddard@xxxxxxxxxxxxxxxxxxxxx using -f
From: Thomas Goddard <goddard att cgl.ucsf.EDU>
To: e.d'auvergne att pgrad.unimelb.edu.au
Hi Edward,
I will try to get the region RMSD feature into Sparky next week.
What platform are you using (Linux, SGI, Windows)?
Tom
"""
and:
"""
Date: Tue, 28 Aug 2001 17:11:32 -0700 (PDT)
Message-Id: <200108290011.RAA331478@xxxxxxxxxxxxxxxxxxxxx>
X-Authentication-Warning: socrates.cgl.ucsf.edu: goddard set sender to
goddard@xxxxxxxxxxxxxxxxxxxxx using -f
From: Thomas Goddard <goddard att cgl.ucsf.EDU>
To: e.d'auvergne att pgrad.unimelb.edu.au
Subject: Re: O.S.
Hi Edward,
I added the spectrum region RMSD command to Sparky 3.101.
The accelerator is rm. It is under the Extensions/Spectrum menu.
There is a short description in the extensions section of the manual.
You can get the Linux version I made for you from:
http://www.cgl.ucsf.edu/home/sparky/prerelease
Hope this does what you need. I added to Sparky's Python interface
the ability to be notified when the user drags a box in a spectrum.
I think this feature will be helpful in future Sparky extensions.
Tom
"""
The problem with the whole spectrum estimate of the RMSD is minimally 3-fold:
1) Water suppression. This varies wildly between experiment types
and suppression techniques (both experimental and processing), and the
noise around the water signal can have a significant effect on this
estimate.
2) Empty parts of the spectrum. Normally you remove part of the
spectrum in the processing as it contains no data. For example the
right hand side of a 15N HSQC. The estimate on the full spectrum and
the extracted spectrum will be different.
3) Baseplane artifacts. If you use Sparky's 'rm' command, you will
see that the noise level changes all over the spectrum. The noise
around your peaks of interest, mainly in the random coil region, will
be higher than the full spectrum estimate. This can be due to a
slight baseplane roll, spectral artifacts, and tiny amounts of
contaminates in the sample.
The 'rm' method allows you to measure the noise around your peaks of
interest and use a much better error estimate. This is especially the
case for peak around the water signal where separate larger errors can
be specified in relax.
In any case, you should always consider your error estimate to be as
important, or even more important than the data itself. The error
hugely influences the optimisation space, incorrect errors can cause
minima to disappear or fake ones to appear, as well as the model
selection (for example for the comparison to insignificance). Using
an incorrect error estimate can be fatal for subsequent analyses.
Regards,
Edward
Re: field strength off-resonance R1rho constant relax time relaxation dispersion