Bruno Vitorge
I just installed the mddnmr2.4.
The computer is running centos 6.6.
During the installation I got the following message:
Thu. 16 Jul. 10:38:51 $ 331: > ./Install
Checks for needed external programs
Done
trying ./binUbuntu64/mddsolver
trying ./binUbuntu64Static/mddsolver
Incident de segmentation (core dumped)
trying ./binXeonE5mkl/mddsolver
trying ./binUbuntu32/mddsolver
trying ./binUbuntu32Static/mddsolver
Incident de segmentation (core dumped)
trying ./binCentOS32/mddsolver
trying ./binCentOS32Static/mddsolver
Install ./binCentOS32Static/mddsolver
Installing for platform CentOS32Static
I'm also using qMDD version 3.2
I was wondering if it is normal that when I select IST algorithm and set the threads to 10 (I have 16 CPU and 32Go of RAM) only one cpu is used.
The region is set to 5, the CS lamda is set to -1.
The other parameters:
#!/bin/tcsh
setenv FID ../1151_ghn_coA_S_16_32
setenv fidSP fidSP.com
setenv REC2FT recFT.com
setenv in_file nls.in
setenv selection_file nls.hdr_3
setenv FST_PNT_PPM 11
setenv ROISW 5
setenv proc_out ft/test%03d.dat
setenv SPARSE y
setenv NUS_TABLE_OFFSET 0
setenv NUS_POINTS 512
setenv NI '16 32 1 '
setenv NIMAX '32 64 1 '
setenv NDIM 3
setenv MDDTHREADS 10
setenv METHOD CS
setenv NCOMP 25
setenv NITER 50
setenv MDD_NOISE 0.7
setenv CEXP nnn
setenv CT_SP nnn
setenv lambda 0.001
setenv SRSIZE 0.1
setenv CS_alg IST
setenv CS_norm 1
setenv CS_lambda -1.0
setenv CS_niter 200
setenv CS_VE y
mddnmr4pipeN.sh 1 2 3 4 5
set ecode=$? ;if ( $ecode ) exit($ecode)
proj3D.tcl -in $proc_out
xyz2pipe -in $proc_out >1151_ghn_coA_S_16_32.ft3
When the MDD algorithm is used several CPU are effectively used.
Bruno
Krzysztof Kazimierczuk
Everything is as it should be - multiple threads are not implemented for IST yet (only in IRLS and MDD). You can however, run parallel calculation, by calling regions.runs instead of steps 2 3 4:
i.e.:
mddnmr4pipeN.sh 1 2
./regions.runs
mddnmr4pipeN.sh 4 5
instead of:
mddnmr4pipeN.sh 1 2 3 4 5
And one more thing: CS_lambda of -1 is very strange setting. It should be either default 1 or something close (like 0.9)
Kind Regards,
Krzysztof
--
--
You received this message because you are subscribed to the Google Groups "mddnmr" group.
To post to this group, send email to mdd...@googlegroups.com rom your registered email address.
To unsubscribe from this group, send email to mddnmr-un...@googlegroups.com
For more options, visit this group at http://groups.google.com/group/mddnmr
---
You received this message because you are subscribed to the Google Groups "mddnmr" group.
To unsubscribe from this group and stop receiving emails from it, send an email to mddnmr+un...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.
Krzysztof Kazimierczuk
Sorry,
Krzysztof
Bruno Vitorge
I will try your solution.
About the CS_lambda value I set it to -1 because want I let the mouse in the box of the help message said -1 (automatic for ist) as you can see on the screen capture:
I'm going to read and hopefully understand the manual, to be able to choose more logically the values.
Also at Euromar in Prague, Orekov presented new algorithms do you think they will be implemented in qMDD?
Thank you again for the fast answer.
Kind Regards,
Bruno
Krzysztof Kazimierczuk
Thanks for the picture - this is actually wrong information. It is CS_norm not CS_lambda, that is automatically switched to 1 (and not to -1) when using IST.
Maxim, could you fix it in qMDD?
Also, I think, that default IRLS algorithm is more efficient, than IST. Unless the datasize is really large, I would try IRLS rather than IST.
Regarding other algorithms - I guess, that Vladislav should answer. From my side, there should be new algorithm LPMP somewhere in the end of 2015 (was published in: Kazimierczuk&Kasprzak, Sensors, 2015).
Kind Regards,
Krzysztof
Krzysztof Kazimierczuk
New version of mddnmr is available at:
http://pc8.nmr.gu.se/~mdd/Downloads/
Several modifications have been made to improve computational time (e.g.
multiple threads for IST) and lessen memory requirments (50% for IRLS).
Update recommended!
Regards,
Krzysztof Kazimierczuk
Bruno Vitorge
and in the archive I found only binary for ubuntu, so when I try the command
./Install
I got the following error messages:
./Install
Checks for needed external programs
Done
trying ./binUbuntu64/mddsolver
trying ./binUbuntu64Static/mddsolver
Incident de segmentation (core dumped)
trying ./binUbuntu32Static/mddsolver
Incident de segmentation (core dumped)
trying ./binCentOS32/mddsolver
trying ./binCentOS32Static/mddsolver
I tried to run with the available binary but it always failed.
The reconstruction seems to be OK, because I got the ft/test%3d.DAT files
but I got :
Thread number 139892185102080 ready for region 32
ft/test%03d.dat ready
Incident de segmentation (core dumped)
Could you provide centos 64 or centos 32 binaries.
Or give me some indications to make mddnmr2.4 work on my system.
Thank you very much for your software.
PS: the mddnmr 2.2 version is not working any more probably because I updated some package (in centos and in Canopy).
Now the GUI start then close with the following error message:
QPixmap: It is not safe to use pixmaps outside the GUI thread
X Error: BadImplementation (server does not implement operation) 17
Major opcode: 20 (X_GetProperty)
Resource id: 0x0
[xcb] Unknown request in queue while dequeuing
[xcb] Most likely this is a multi-threaded client and XInitThreads has not been called
[xcb] Aborting, sorry about that.
qMDDGUI.py: Fatal IO error: client killed
python: xcb_io.c:165: dequeue_pending_request: Assertion `!xcb_xlib_unknown_req_in_deq' failed
I was trying to convince my co-workers to use NUS but as qMDD is not working any more on our computer I'm a bit embarrassed :-(
Krzysztof Kazimierczuk
Vladislav has just made a new version of mddnmr that should work on Centos. Please go to NEW DOWNLOAD WEBPAGE and download mddnmr2.4:
http://mddnmr.spektrino.com/download
Kind Regards,
Krzysztof
--
--
You received this message because you are subscribed to the Google Groups "mddnmr" group.
To post to this group, send email to mdd...@googlegroups.com rom your registered email address.
To unsubscribe from this group, send email to mddnmr-un...@googlegroups.com
For more options, visit this group at http://groups.google.com/group/mddnmr
---
You received this message because you are subscribed to the Google Groups "mddnmr" group.
To unsubscribe from this group and stop receiving emails from it, send an email to mddnmr+un...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.
-- Pozdrawiam, Krzysztof Kazimierczuk<br /> Krzysztof Kazimierczuk, PhD Laboratory of Intermolecular Interactions University of Warsaw Zwirki i Wigury 101, 02-089 Warsaw tel. +48 22 8220211 ext. 421
Bruno Vitorge
Thank you very much for this new version which seems to work perfectly.
Kind Regards,
Bruno
Vladislav Orekhov
Dear MDDNMR users,
We like to announce a set of examples designed for comparing performance of several popular NUS processing methods from the MDDNMR package and other programs: MDD [1], IRLS [2], IST[2], LR[3], hmsIST [4], NESTA [5], SCRUB [6], and SMILE [7].
We start with representative protein spectra 2D HSQC, 2D NOESY, 3D 1H-15N NOESY-HSQC, and 3D HNCO, recorded in full. MDDNMR allows to sparsify full spectra for a given NUS schedule. All algorithms use exactly the same input and output in nmrPipe format. Thus a fair comparison of different algorithms and NUS schemes in respect to the reconstructed spectra quality is possible.
We are looking for repeating the procedure for many sampling schedules to make the analysis more representative. However, we like to make the scripts and examples public as soon as possible in order to allow the discussion and independent testing.
The processing script and analysis used in the examples are general and can be used for comparisons using any other full spectrum. For the processing, the scripts use default/recommended parameters. You can look at (e.g. in nmrDraw) and compare the reconstructed spectra with the reference and between different algorithms. In addition, there is a simple analysis of the spectra quality. Namely, intensities of the peaks defined in the nmrPipe peak lists are compared in the reference and reconstructed spectra. The correlation plots and correlation coefficients are displayed in gnuplot.
In order to process the spectra and look at the comparisons results follow the procedure described at http://mddnmr.spektrino.com/comparisons
[1] V. Y. Orekhov, V. A. Jaravine, Prog Nucl Mag Res Sp 2011, 59, 271-292.
[2] K. Kazimierczuk, V. Y. Orekhov, Angew. Chem.-Int. Edit. 2011, 50, 5556-5559.
[3] X. B. Qu, M. Mayzel, J. F. Cai, Z. Chen, V. Orekhov, Angew Chem Int Edit 2015, 54, 852-854.
[4] S. G. Hyberts, A. G. Milbradt, A. B. Wagner, H. Arthanari, G. Wagner, J. Biomol. NMR 2012, 52, 315-327.
[5] S. J. Sun, M. Gill, Y. F. Li, M. Huang, R. A. Byrd, J. Biomol. NMR 2015, 62, 105-117.
[6] B. E. Coggins, J. W. Werner-Allen, A. Yan, P. Zhou, J. Am. Chem. Soc. 2012, 134, 18619-18630.
[7] J. Ying, F. Delaglio, D. A. Torchia, A. Bax, J Biomol NMR 2016, 1-18.