Exercises NMRPipe data processing of 2D and 3D NMR data These Exercises NMRPipe by pengtt


									Exercises NMRPipe data processing of 2D and 3D NMR data

These exercises assume that a recent version of the NMRPipe package is
installed on your system, and should work irrespective of the platform
you use. For downloads, know bugs, documentation, demos and reference
material please refer to http://spin.niddk.nih.gov/NMRPipe or the mailing
list http://groups.yahoo.com/subscribe.cgi/nmrpipe.

In the directory "NMRCourse" you will find three Varian datasets:


Each of these data directories is wrapped up in a tar file, and compressed.
Download them to your computer. The data can now be extracted in the
following fashion:

zcat hnco.tar.Z | tar xvf

The purpose of the 2D datasets is to illustrate the use of functions and
parameters in NMRPipe. Acquisition parameters are contained in the file
"procpar", and can be extracted with the auxiliary "varian" conversion
utility supplied with the NMRPipe suite. For Bruker data a similar tool
exists. Be aware that these scripts extract parameters based on a likely
implementation and execution of the experiment. The utility will guess
the order and associated frequencies of the n-dimensional experiment.
Also it will assume that the carrier frequencies were tof, dof, dof2, etc
(Varian) or O1, O2, O3, etc (Bruker), and not modified by changing the
carrier during the experiment (either in the pulse sequence, or by a
parameter associated with it).

The HNCO dataset is used to illustrate conversion and processing of 3D
data, again including the use of linear prediction.

Good luck with the exercises!

Frans Mulder
University of Groningen, May 2008

The purpose of this exercise is to familiarize your self with processing
from the UNIX shell command line, as well as with the nmrDraw

The dataset was obtained using a gradient sensitivity enhanced 2D 15N-
1H HSQC experiment (pulse sequence from the Toronto library of Lewis

1) conversion
Start the program "varian" from the shell inside the .fid directory. Us e
Read parameters to extract standard parameters from the procpar file.
Then check the parameters, and adjust those that you think - or already
know - were not interpreted correctly. If you miss anything at this point it
may become apparent later, and you can return to this point. When you
are satisfied with the parameters select Save script and Execute script.

Alternatively one can use from within nmrDraw File>Macro Edit to
create a conversion script under Varian Conversions>2D conversion.
This will then require you to know which parameters to look for in

2) checking your time domain data
Start nmrDraw by calling it from the shell command line. Be aware that
items from the pulldown menus are selected by using the right mouse
button. Select the converted FID, and read/draw it to inspect the data. For
example, have a look at the first FID by use of Mouse>1D horizontal. Is
everything as expected?

If the conversion went well the time domain dataset consists of
(512x2)x(128x2) points. Realize that you are now looking at only half of
the data! You only see the "real" FIDs in the acquisition dimension, i.e.
collected and compared with a cosine carrier wave. The imaginary FIDs
are hidden to your view. If you want to look at one of them then you can
select this FID and choose: proc>NMRPipe command. On this
command line you can type "SHUF -swap". Now you are looking at the
imaginary part of the data.

Sometimes there may be corrupted FIDs, for example due to receiver
overflows. So it is important to be able to inspect all the time domain
data. Using macro edit write a small processing script that reads in the
data; shuffles the real and imaginaries; writes out the data file. Now you
are ready to inspect the hidden part of your data.

3) Processing from nmrDraw
Choose Proc>NMRPipe command for the processing popup window.
Using the first FID perform a sequence of 1D processing commands to
obtain a frequency domain spectrum. You may start over when trying the
effect of another command, or changing some paramaters (e.g. window
function). If you would like to visualize a window function you can type
"set �r 1" when a sequence of points (an FID ...) has been read in. This
will replace all the points of the original vector with ones. If you now
apply the function of the desired function you can inspect the result on the
screen. For example, try a Lorentz-to-Gauss transformation and see how
the parameters influence the resulting window function. Select a window
function that enhances the resolution of your data, with little ripple, that
can be used for 2D processing. Determine the phase corrections from the
nmDraw window. For this you must first activate the phasing tool!

4) Creating an interferogram
Within the File>Macro Edit window select Process 2D>Basic 2D. Edit
the macro to incorporate the processing of the acquisition domain.
Comment out processing of the indirect domain by placing # at the start
of any lines associated with this dimension. Add additional Windows,
Transforms, Corrections where needed. Save the script with an
appropriate name when ready. Use the help button to get information
about a function and its arguments. Execute your script and inspect the
result: Select a vertical trace to try out the processing parameters in the
indirect dimension. Phasing parameters are often known before hand,
depend on how much evolution has taken place when the first point was
measured. In the present experiment acquisition started at zero.
Remember what that means for the first point?

5) Processing the 2D spectrum
Now complete the 2D processing script with all operations you want
included, and execute it. Check the result. If you notice any undesired
features then go back to the conversion or processing scripts and then edit
and rerun these.

6) Comparison with a reference spectrum
Below is the final spectrum I obtained. How did you do? If your spectrum
contains mirrored peaks in the 15N dimension, and it appears impossible
to phase it means that you have set -yMODE to Complex in fid.com.
However, since this data was obtained with sensitivity enhancement,
employing field gradient coherence selection you first need to make
linear combinations of the input data. This is done implicitly by setting -
yMODE to Rance-Kay. To learn more about this procedure please refer
to recent textbooks on (biomolecular) NMR spectroscopy, or the original
paper: L.E. Kay, P. Keifer, T. Saarinen, J. Am. Chem. Soc. 114, 10663-
10665 (1992).

This is an example of constant-time data in the indirect domain. The
experiment was recorded on a 50% deuterated sample, as you will notice
in a short while. The purpose of the exercise is to process truncated data
correctly, and to use linear prediction.

1) conversion
Start the program "varian" from the shell inside the .fid directory and
convert the varian fid file to NMRPipe format. Process the acquisition
dimension. This may be difficult.

2) Process the indirect dimension
Select a trace in the t1 direction. Convince yourself that this is really
constant-time data. Find an appropriate apodization function. Think of
S/N, resolution, ripple. Can you phase this data? Can you rationalize the
phasing parameters? How do you think I decoupled the carbonyl carbons?

3) Linear prediction
Now extend the original t1 data by linear prediction. There are various
options for this in the package. Which would be appropriate? Does the
resolution increase? How would you adjust your apodization on the
extended data? How does this influence your S/N?

4) 2D processing
Process the 2D to your satisfaction. Do this once with LP and once
without. Compare the two spectra as whole. How does it look? Now
zoom in on area with a few peaks in a crowded region. Look carefully at
both spectra. What do you think now? Can you explain the peak shapes in
the spectrum that used LP?

1) conversion
Start the program "varian" from the shell inside the .fid directory and
convert the varian fid file to NMRPipe format. The way that 3D data are
recorded is usually to acquire the complex points in the indirect
dimensions consecutively. There are two possibilities for this for 3D data.
In var2pipe the order is indicated by the parameter aqORD:

-aqORD aqOrd [0] Acquisition Order Code.
  0 = 3D d3,d2,phase2,phase
  1 = 3D d3,d2,phase,phase2

Check procpar to know which option is appropriate. Convert the 3D.

2) 2D plane processing
Under Process 2D you can find scripts to process the XY and XZ planes
of the 3D dataset. Edit to your liking and process the two planes. Check
how LP may be implemented in each of the indirect domains.

3) 3D processing
Using the parameters optimized for the 2D plane processing build a 3D
processing script with help from File>Macro Edit, excluding LP. This
should be quick. Make sure you are happy with the results. For example,
are the frequency axes running in the correct sense?
Next we will try to implement LP in both indirect dimensions. This will
take longer so inspect your scripts well before execution. Propes use of
LP requires that it is applied to a dataset that has only the dimension to be
predicted in the time domain, and all other dimensions processed. This
will require the use of inverse functions as well. Check in File>Macro
Edit>3D processing.

Finally transpose the 3D spectrum without LP so that you can now
inspect both (H,N) planes as a function of 13C' (z), as well as (H,C')
planes as a function of 15N (z). Unfortunately nmrDraw does not allow to
"role the axes".

Export the result to your favourite analysis program (Sparky, NMRview,
Additional remarks and questions to the n otes of Cornilescu

Before you "process the data in the usual way" have a look at the fid.com
and ft2.com scripts and convince yourself that you understand them.

After you have gone through this chapter note the following: To a slightly
experienced NMR spectroscopist the 2D spectrum appears "funny". What
are those signals that appear near 145 ppm in the nitrogen dimension?
Are there any nitrogen nuclei in a protein that resonate in this range?

Apropos, around which chemical shifts should you expect signals of (a)
backbone amide (b) histidine protonated and deprotonated side chain (c)
arginine N and N (d) the N-terminal -amino group (e) lysine N (f)
asparagine and glutamine amide (g) tryptophan indole nitrogen nuclei?

Reprocess the spectrum with the nitrogen frequency axis running in the
opposite direction by changing the argument to the function �fn FT.
What are those signals around 85 ppm due to?

Before you start picking peaks make sure that phasing is already perfect.
Then adjust the script to pick only positive peaks. How many peaks are
there in the HSQC spectrum? And divided into how many peak clusters?

After processing the ipap spectra have a look at the level at which you see
"dif" peaks in the "sum" spectrum. Have a look at the included
coad_1234.M macro and see if you can improve this.

After processing the data inspect the scripts that are called by fit.com.
After running fit.com we will not proceed with model.com, but run the
interactive script showEvolve.tcl and fit some curves interactively.

Next, change the processing script to use matched filters in F1/F2. Inspect
the difference spectrum (between the input and simulated data). How
good is the fit? In which way does your choice of window function
change the goodness of the fit? And how the completeness of fitted

To top