Problem with MDD processing

[kai xue]

Dear friends in the group,

Hello, I have some problem with MDD method. I want some advice for my HSQC 2D (25%) and H-C-C 3D (10%) NUS experiment.

The measurement is performed with Bruker 400 HMz. data is processed with topspin 3.2 and qMDD 2.0 respectively.

When I process 2D and 3D data with CS option, it works.

However when I processed data with method option of MDD, there is no peak on the spectrum. I also used topspin and qMDD both, it is not there.

I want some advice on parameters setting for MDD method( For example NCOMP, RMDD, NOISE,  NITER, etc.). 

Experimental data are included in the attachment, it will be appreciated if there are some suggestions of how to set MDD processing.

Thanks.

Yours sincerely

Kai

[Vladislav Orekhov]

Dear Kai
I looked at your 2D spectrum in qMDD. First, Bruker experiment directory must be a number for qMDD to work properly, so I renamed 2D dir to 2. Then, most of the signals are found around -5 ppm in 1H, you need to set parameters FST_PNT_PPM  and ROISW properly.
Third, the NUS spectrum contains 16 complex fids out of total 64 in the fully sampled spectrum. This is a bit too little for a good reconstruction.

Indeed, CS IRLS algorithm works best for this spectrum. MDD works as well, but resulting spectrum is worse. Below are parameters, which I used for CS and MDD, these you can find and edit in file proc.sh (qMDD). 

setenv FST_PNT_PPM 10
setenv ROISW 25
setenv MDDTHREADS           5
setenv METHOD               MDD
setenv CS_alg               IRLS
setenv CS_norm              0
setenv CS_lambda            1.0
setenv CS_niter             20
setenv SRSIZE               0.2
setenv NCOMP                12
setenv NITER                250
setenv MDD_NOISE            0.7
setenv CEXP                 yn
setenv CT_SP                nn
setenv lambda               0.01

Best,
Vladislav

--
--
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.

[kai xue]

Dear Vladislav,

I tried process my 2D and 3D data with qMDD again. Same problem remains, MDD does not show all the peaks after reconstruction, but CS gives a much better result.

Then my questions are,

1 For NUS measurement, is there a limit for CS and MDD algorithm? For example, as you said, 2D measurement 25% sample CS works better. Is there a similar sampling rate limit for 3D and 4D NUS measurement?

2 In topspin, MDD algorithm works up to 5D, CS does not work for more than 3D. Is there a similar limitation for qMDD?

Thanks!

Yours sincerely

Kai

在 2014年8月12日星期二UTC+2下午4时21分45秒，Vladislav Orekhov写道：