Hi Martin,
This is a difficult question to answer. From memory, linear
prediction has been mentioned in one or two of the hundreds of
model-free papers published to date. However I cannot remember which
papers these would be. And it is probably just a single line in the
back of a discussion somewhere. By searching through my collection of
papers, I found the following more detailed reference:
N. J. Skelton, A. G. Palmer III, M. Akke, J. Kördel, M. Rance, and W.
J. Chazin, J. Magn. Reson. B 102, 253 (1993).
This is probably the most detailed study of linear prediction in NMR
relaxation, but I would not call it comprehensive. As far as I am
aware, there is no systematic study on the effects of linear
prediction on a dynamics analysis. For example what happens with
different levels of spectral data truncation, different number of
linear prediction coefficients, linear prediction in different
dimensions, as well as some of the other features of linear prediction
not implemented in NMRPipe. It is well known that linear prediction
can introduce artifacts, however how this translates into relaxation
data or, more importantly, the model-free parameters is completely
unknown. The problem with dynamics in NMR is that this requires the
highest precision and highest quality data possible - far greater than
any other NMR technique. And therein lies the problem - without a
comprehensive study of how linear prediction affects the final
dynamics, you can never know what problems or artifacts that might
introduce. And such artifacts may not be distinguishable from real
results. Such a study could probably be published as a standalone
paper.
Anyway, you should probably look at performing the same types of
testing as in the above reference if you would like to get into linear
prediction. You should also try processing without window functions,
as well as processing without the Fourier transform in the indirect
dimension to understand the level of truncation you have in the base
data. Then if the Lorenzian to Gaussian window function amplifies the
truncation too much, then it should be dropped. I usually use the
NMRPipe GM in the direct dimension and the 60 degree shifted sine
squared bell in the indirect, as I mention in the relax manual
(http://www.nmr-relax.com/manual/Spectral_processing.html). If you
have truncation in the direct dimension and are not working with small
organic molecules, something is strange (even with small molecules it
would be strange). You can see this if you process without any window
functions, linear prediction, baseplane corrections, and the Fourier
transform in the direct dimension. I hope some of this helps.
Regards,
Edward
On 19 March 2014 15:53, Martin Ballaschk <ballaschk@xxxxxxxxxxxxx> wrote:
Hi Edward and relax-users,
I was again thinking about the ideal processing strategy of my R1/R2
relaxation and HetNOE planes. Routinely in the past, I used the Topspin
Sine-Bell functions and linear prediction for resolution enhancement in the
indirect dimension. Now, I use NMRpipe with Lorentz-to-Gauss windows, but
with my collected data, I hardly get the needed resolution without having
severe truncation artifacts from my strong peaks wich contaminate
neighbouring peaks' intensities.
The latter issue is why I don't use two sets of processing parameters (one
for intense and one for weak peaks), but fiddle around with overlapping
peaks.
I know that you advise to avoid linar prediction, but after reading
http://spin.niddk.nih.gov/NMRPipe/ref/nmrpipe/lp.html I have the impression
that LP could help ease the problem of truncation artifacts. I also did
some literature searching, but I didn't find anything about LP making peak
height measurements unreliable.
I remember we discussed that during your visit, and I showed you the graphs
where I compared rates calculated from spectra without LP vs. rates from
spectra where the same number of points was added by LP. Maybe you
remember, there was no visible bias, but rates with large errors also
became larger.
So what again is the reason to not use LP for relaxation series?
Cheers,
Martin
_______________________________________________
relax (http://www.nmr-relax.com)
This is the relax-users mailing list
relax-users@xxxxxxx
To unsubscribe from this list, get a password
reminder, or change your subscription options,
visit the list information page at
https://mail.gna.org/listinfo/relax-users
Re: Processing again