The MDDNMR web server tutorial

created by Victor Jaravine, Bio.Mag.Res.Zentrum, Frankfurt, Sep 2010

 

The MDDNMR portal

 

The portal performs reconstruction and processing of Non-Uniformly Sampled (NUS) NMR experiment data using the eNMR server: http://www.e-nmr.eu/webportal/. The server requires registration (free for all academic users). In addition, using the GRID high-performance computation infrastructure requires a valid personal certificate. For information on how to obtain such a certificate please refer to: http://www.enmr.eu/eNMR-registration

 

Required software for local viewing of results

                 nmrDraw (is part of nmrPipe package http://spin.niddk.nih.gov/NMRPipe/ ), or any other spectra visualization software that can read nmrPipe format data, for example, CCPN http://www.ccpn.ac.uk/ccpn/software/ccpnmr-suite, CARA http://cara.nmr-software.org/downloads/ etc.

Relevant web sites

http://groups.google.com/group/mddnmr/

http://www.enmr.eu

 

 

I. Overview

In this tutorial you will perform reconstruction and processing of an example of Non-Uniformly Sampled 3D NMR experiment data using the eNMR GRID server.

 

Conventionally, MDDNMR performs calculation with most parameters set at default values, and there is no need to learn them and change the defaults without a specific purpose. A dozen or so of parameters can be changed via the portal web-form, but in this tutorial we shall only change a few.

 

There are 5 advanced examples of NMR experimental data of varying spectral dimensionality; these data were used in the corresponding publications (the details of the advanced datasets is given below):

 3D HNcoCA


 2D HMQC


 4D NOESY


 3D 15N NOESY-HSQC


 3D HNCA

But in this tutorial we provide step-by-step description only for the 1^st example 3D HNcoCA.

 

The examples can be also run in UNIX line command mode, using standard gLite User Interface (UI) on the eNMR GRID. For this method of processing you will need an installation of gLite UI on your computer or notebook, and a couple of other files needed for submission, the files are listed and the procedure is described on the eNMR wiki, titled as use-case for MDD http://www.e-nmr.eu/use-case-mddnmr

 

Here we focus only on the mode of submission via the web-portal using a standard web browser (Firefox recommended).

 

 

II. Job submission

Required input data

In order to run this tutorial, download the data by following link webmdd_hncoca_fid.tar  

Unpack this file in a directory of your choice:

tar xvfz 284hncoca.tgz

This will create three input files for running of the example via web-portal:

a tar-gzipped pack of original FID data (recorded by a NMR spectrometer in NUS mode) 284hncoca.fid.tgz , and two conventional [=DFT] processing scripts for nmrPipe (by some estimates 70% of all spectra are processed with this software) fidSP.com and recFT.com, located in a directory 284hncoca.proc.

 

Submitting the data via the web-form

At http://www.enmr.eu/webportal/ go to MDD http://www.enmr.eu/webportal/mdd.html

Uploading experimental data

The first file to upload is the .tgz (tar-gzip) archive (it can be also simple .tar) containing experimental data: 284hncoca.fid.tgz (containing FID directory 284hncoca.fid with a binary data FID file, in this case Bruker ser (or can be Varian fid), and all associated experimental parameters inside the .fid directory (note, that we renamed and added extension .fid to Bruker digits name, that is done for clarity of presentation and it is also Varian convention.

 

Uploading nmrPipe scripts

Upload the text file scripts without any modification:

fidSP.com - the script is used for processing of direct dimension of the fid, (it is fully sampled as this is direct dimension); it is invoked prior to running mdd reconstruction of the indirect dimensions. This file is automatically generated by Varian vnmr software.

 

#!/bin/csh

# : x1 xn data B xphase

#echo $0 $1 $2 $3 $4 $5

set xa = $1; set xb = $2; set data=../{$3}.data;

set name=$4; set xphase=$5; set fid=../{$3}.fid/fid

 

bruk2pipe -in $fid -bad 0.0 -aswap -DMX -decim 2773.33333333333 -dspfvs 20 -grpdly 67.9858856201172 \

 -xN 1024 -yN 2 -zN 720 \

 -xT 512 -yT 1 -zT 360 \

 -xMODE DQD -yMODE Complex -zMODE Complex \

 -xSW 10000.000 -ySW 2500.000 -zSW 2500.00 \

 -xOBS 600.1328168 -yOBS 150.910829 -zOBS 60.817688 \

 -xCAR 4.725 -yCAR 55.843 -zCAR 115.860 \

 -xP0 -213.5800 -yP0 0.00 -zP0 0.00 \

 -xP1 26.200 -yP1 0.00 -zP1 0.00 \

 -xLAB 1H -yLAB C13 -zLAB N15 \

 -ndim 3 -aq2D States \

 -verb \

# | nmrPipe -fn POLY -time -auto \

| nmrPipe -fn SP -off 0.330 -end 0.970 -pow 1 -c 0.500 \

| nmrPipe -fn ZF -auto \

| nmrPipe -fn FT -auto \

| nmrPipe -fn PS -hdr \

| nmrPipe -fn PS -p0 ${xphase} -p1 0 -di \

| nmrPipe -fn EXT -x1 ${xa}ppm -xn ${xb}ppm -sw \

| pipe2xyz -z -out ${data}/${name}%03d.DAT -ov -nofs -verb

exit 0

 

and recFT.com - this file makes DFT transform of the two indirect dimensions after the reconstruction procedure (it can used without any change for any 3D spectra).

 

#!/bin/csh -f

echo '##############' in $0 $1

if( $#argv < 1 ) then

echo Use: $0 <input pipe> <template for output spectrum>

echo nmrPipe processing of YZ dimensions after MDD reconstruction

exit 1

endif

 

set ft4trec=$1

if( $#argv > 1 ) set proc_out=$2

 

if( ! -f $ft4trec ) then

 ls $ft4trec

 echo $0 failed

 exit 2

endif

 

echo '######### Processing time domain MDD reconstruction #############################'

echo

echo Processing YZ dimensions

showhdr $ft4trec

 

cat $ft4trec \

# XYZ \

| nmrPipe -fn TP -auto        \

# YXZ \

| nmrPipe -fn SP -off 0.45 -end 0.98 -pow 1 -c 0.5 \

| nmrPipe -fn ZF -auto \

| nmrPipe -fn FT                                                  \

| nmrPipe -fn PS -hdr \

| nmrPipe -fn PS -p0 0 -p1 0 -di \

| nmrPipe -fn TP -auto                                     \

# XYZ \

| nmrPipe -fn ZTP                   \

# ZYX \

##| nmrPipe -fn LP -auto -ps0-0 \

| nmrPipe -fn SP -off 0.45 -end 0.98 -pow 1 -c 0.5 \

| nmrPipe -fn ZF -auto \

| nmrPipe -fn FT        \

| nmrPipe -fn PS -hdr \

| nmrPipe -fn PS -p0 0 -p1 0 -di \

| pipe2xyz -out $proc_out -x -ov

# ZYX

 

echo $proc_out ready

exit

 

The scripts of course can be changed for any required change in processing, e.g. one can change window function, phases or uncomment step for with mirror image linear prediction for Z dimension above:

| nmrPipe -fn LP -auto -ps0-0 \

 

The runs:

Generally, two runs for comparison of the reconstruction quality; but if you believe us that noise in HN plane of sparse-DFT is some 100x higher compared to the MDD reconstruction , and as it takes ~30 mins for each run by the GRID even when job traffic is low, then we suggest you to do only Run2:

Run1: check option for sparse-DFT processing – in this case the missing time-domain data points are set to 0, followed by regular 3D DFT (i.e. without MDD).

Run2: check option for MDD processing – in this case the missing time-domain data points are reconstructed by R-MDD procedure, followed by regular 3D DFT.

 

Change other parameters on the web-form

The parameter that is usually changed is ROI size (ppm) (ROI - Frank Delaglio s nmrPipe term denoting Region-of-Interest ). Change it from 0.45 to 1.0 for example, (or higher 2.0 or 3.0 ppm). ROI is the size for the extract from the direct dimension. Thus with leftmost point set at 8.75 , the region of interest 1.0 corresponds to the ¹H^N range 8.75-7.75ppm.

Important: Check the option Bruker for this example. Currently MDDNMR uses two naming notations for the root name of the NUS files: nus_3dsetup (Bruker) and nls (Varian).

Job submission

As a final step, enter your username and password and press Submit. If everything went correctly, MDDNMR will provide you with a link to the result. The page will check if you entered your parameters correctly. If error has occurred during calculations, it will be in the output std.err file.

 

III. Viewing the results

Since we are calculating a 1ppm extract of a relatively small 3D spectrum, the MDDNMR runs should take no more than 1 minute, or more depending on the server current load. However, due to some accounting requirements imposed on the EU-funded (Brussels) project, each job needs to wait at least 30 mins at a server Florence for registration.

Once the run is finished, it is Ok to unpack the resulting file, and check the results. Just in case, we have the pre-calculated results, which can be retrieved using the job number 1313666300pYC8xf

 

Visual inspection and comparison

Visual inspection of the results is an important part of the analysis. You can view the quality of reconstructed and DFT processed spectra - by viewing in nmrDraw (or other program like CCPN, which is installed on the current system) the 3D itself, or by viewing three 2D projections *.dat files, or simply view these printed as *.ps file.

 

If you performed two runs, unpack them into different locations (or rename the first downloaded directory), and compare the spectral quality. The most noise in the sparseDFT case is in 2D H-N plane (N15.1H.dat), that is very noticeable difference to close-to-zero noise in the mdd run, the other sparseDFT planes are quite noisy as well.

You could check this fact quantitatively by running a noise estimation procedure (e.g. Estimate noise in the Draw menu of nmrDraw).

 

IV. (Optional) additional runs

After the initial experience you could try other examples or your own experimental data. The examples can be executed on the eNMR GRID in exactly the same way as the above example (3D HNcoCA), using the web-form submission/retrieval, but different tarred FID and the different nmrPipe scripts. Several more examples, described in the documentation can be downloaded directly from site of SNC-Hasselblad-lab. (Please, note that some files have large size).

 

Alternatively, you can run MDDNMR on your own data, if you have recorded experiments in NUS mode. In the latter case, one need to upload a corresponding tarred fid, and use fidSP.com located in the *.fid/proc/ directory (it is automatically created by VNMR software on Varian; for Bruker this is done differently, but the easiest explanation at the moment is to say that such file can be prepared via manual copy-paste parameter header from the file fid.com produced by running the program bruker from the nmrPipe package, to replace corresponding lines in fidSP.com). For any 3D case recFT.com can be used without modification. Alternatively, especially for higher dimensions, the scripts can be prepared manually by modification from the one of the examples of the corresponding dimensionality.

 

 

3D HNcoCA: 284hncoca.tgz (6.0 Mb), Ubiquitin (Bruker data)
Jaravine V, Zhuravleva A, Permi P, Ibraghimov I, Orekhov VY. J. Am. Chem. Soc. 2008, 130:3927-36

 

2D HMQC: HD384_plasma_gChsqc.tgz (11.7 Mb) VDAC (Bruker data)
Hiller et al. Science 321, 1206-1210 (2008)

 

4D NOESY: A_63_VDAC_25demo.tgz (21.0 Mb) VDAC (Bruker data)
Hiller et al. Science 321, 1206-1210 (2008)

 

3D 15N NOESY-HSQC: BPgnoesyNhsqc_S.tgz (93.5 Mb) 15 kDa (Varian data)
Orekhov et al. J. Biomol. NMR 2003, 27, 165

 

3D HNCA: az_HNCA_high_res.tgz (845 Mb) Azurin (Varian data)
Jaravine V, Ibraghimov I, Orekhov V. Nature Methods, 2006, 3: 605

 

 

*** ***