AMMRL: Re: Contour peak phasing issues in hsqc spectra -- resolved thanks!

From: John Ralph <jralph_at_wisc.edu>
Date: Mon, 27 Nov 2006 01:05:38 +0100

Dear AMMRL:

Wow, am I impressed with this group. Thanks so much for your responses.

Most had to do with using adiabatic pulse sequences to combat the J-
mismatches. Since these are large, fast-relaxing polymers, we used
Bruker's hsqcetgpsisp.2 pulseprogram (rather than hsqcetgpsisp2.2
which is better for small molecules). Frankly, the results were
ASTOUNDING. I'm convinced, and we'll probably never use the plain old
HSQC again!! I don't know why I thought I had trouble understanding
this before, but it was trivial to set up using Bruker's recommended
parameters. p.s. This being a cryoprobe system, we also did just
comment out the p28 line -- as recommended by Bruker, this seems not
to be necessary.

Thanks so much again. Here, for your information, is a summary of the
many suggestions. I'm not sure of the policy so I left the
submitters' names off here, even though I'd really like to credit
them for their wonderful and helpful responses.

Best regards, and thanks to all again.
John

-------------------
I can talk a bit about the problem you are seeing. It has been
observed and discussed, if I understand what you are describing
right, in the literature. In particular, mismatches between J-
couplings and the delays in an HSQC can cause such phase problems.
Frequency swept adiabatic 180s, such as implement in the CRISIS
sequence of Krishnamurthy and co-workers (I'm at home, and don't have
the reference handy, but will forward it on when I get back in at
work), are one solution to this problem. Kay and also ?? had
original publications discussing this type of solution to the
problem. Drop-off in bandwidth coverage for 180 pulses at very high
field can also produce phase distortions and intensity loss, but
primarily at the edges of the spectrum in the 13C dimension.
Standard adiabatic pulses will cure this problem.

On our Varian spectrometers, I have implemented Krish Krishnamurthy's
HSQCAD sequence, which incorporates frequency-swept adiabatic 180s,
and the improvements in the data quality are dramatic for DEPT-135
analog HSQCs (where -CH2- are inverted from -CH3 and >CH-). The
improvements are apparent for all same-phase data as well, but not as
obvious. I installed the sequence on the 900 in NMRFAM a couple
months ago, and it gave similar improvements there. But it is
important even at much lower field because of the problem with J-
coupling mismatch, which is field independent.

So you are headed in the correct direction with adiabatic 180s, but
the frequency sweep is important for roughly correcting for 1-bond
13C-1H variations in J-coupling values. I am fairly certain that
Bruker has such sequences available; I've seen a couple talks by
Bremel(?) about these implementations in the Bruker pulse sequence
library.

For all that, I have to say that I still see some phase problems in
HSQCAD data on occasion, especially it seems for methoxy methyls, for
example. And I really don't know quite why. On a couple occasions,
I was careful to make sure everything was calibrated, and delays were
long enough, and still saw the phase problems. Your question is a
good one for AMMRL, and I would be very interested in seeing a
summary of the responses you get.
----
I would definitely try hsqcetgpsisp.2 (all 180's replaced with shaped
pulses).
In my experience a nonzero p28 never helps anything, and in the case of
samples in water, it is a giant leap backwards.
----
John - I am assuming these are H-C HSQC's? I think you will see a big  
improvement if you use either the si2 sequences which use gradients  
in the back inept to clean things up and/or use adiabatic 180's on C,  
especially at 750 but also at 500 too. I use them here on our 500 and  
I also do adiabatic 13C decoupling rather than garp. I am not in  
front of the spectrometer but I believe these are the sp.2 variants  
of hsqc. The adiabatic decoupling is great but you may have troubles  
with decoupling sidebands so be careful.
----
Several things come to mind as potential causes for this problem.
1) What software are you using to process your data? NMRPipe fails to  
recognize that Varian gNhsqc data is acquired with sensitivity  
enhancement, and require Rance-Kay F1 phasing instead of States or  
Complex. Could your offline processing software fail to recognize  
this in your Bruker data?
2) I've encountered oddly-shaped gNhsqc crosspeaks when the  
refocusing gradients aren't calibrated just right. On a Varian, this  
means gzlvl1 should be fixed, and gzlvl2 should be optimized to yield  
maximum intensity. Unfortunately, this can't be fixed in processing.
3) Lock phase could be off, which would also require re-acquisition.
Hope this helps. I'll be interested to see how you resolve the problem.
----
I've encountered some phasing problems with HSQC experiments before,  
and it sounds like you have been on the right track when dealing with  
them.  But for what it's worth, here is my take on it (in no  
particular order):
invietgpsi  ==>  hsqcetgpsi  is the new Bruker name for that pulse  
program.
I'm a little surprised you are seeing the phasing issues on several  
different spectrometers and different fields.  It sounds to me that  
it could be an imperfect pulse issue.  Using adiabatic pulses for  
inversion and refocusing 13C pulses would be a very big improvement.   
In fact, at 500MHz and above, I have seen marked improvements when  
using adiabatic pulses versus the standard hard pulses.  I would  
recommend you check out the following pulse program from the Bruker  
library:
hsqcetgpsisp2.2
The pulse program gives suggested pulses for inversion and  
refocusing.  At 750MHz, you should probably use the crp80 variations  
instead of the crp60.  I'm sure someone at Bruker can provide more  
detailed advice on that.
I'm not quite familiar with the old version of the pulse program, but  
the new one calculates the d4 delay based upon the value of the one- 
bond 1H-13C coupling constant that you provide.  I think the default  
setting is 145Hz, which is a compromise.  Depending on your  
particular molecule, you may need to play around with that a little bit.
Regarding the prescan delay time, because you are using digital  
acquisition,  it is not as critical to set "DE" as it was with  
analogue systems.  To put it another way, I have generally found it  
best to leave DE=6us and let the digital filters deal with it.  I  
have found that when DE was set too long, the phasing became a real  
issue, even for a simple 1D proton-NMR experiment.  So check the DE,  
and if it has been changed, then set it to 6us.
Of course, there is always the possibility that this could be a  
processing issue.  But this may be best handled by looking at a  
dataset.  Can you make a dataset available for download?
----
We have seen this effect in protein HSQCs where the intense (AND  
narrow) peaks can not be quite phased simultaneously with all other  
lower intensity (and broader) peaks. Although I have not tried to  
prove this experimentally, I am almost certain that this is caused by  
the "leakage effect" in discrete Fourier Transformation. See the very  
useful book by  Hoch and Stern: "NMR Data Processing" - pages 25 -26.
In your case it could be quite a nuisance. The remedy will be  
increasing the number of acquisition points. It may not completely  
eliminate the "phase error" but should reduce it.  When there are  
only a few of such peaks (as is the case for proteins) one might be  
able to find a position for the carrier such that the inverse of the  
frequencies of the offending peaks is an exact multiple of the dwell  
time.
Let me know what you find if you are going to experiment with the  
number of points or the carrier frequency.
----
I have all Varian, but I thought I would suggest a couple things:
1.  Use the Bruker 2D test sample that they use for installation  
checks.  Varian sends out an indanone sample and we routinely use it  
for troubleshooting and to verify that the 2D exps are working  
properly.  If it works the way it should then the spectrometer is  
fine and so is the processing macros you are using.
2.  Do you check your proton pw90 on the sample after tuning?  There  
are samples that the tuning of the probe will still not give you the  
same value of pw90.
----
It would be very nice to see a software program be able to phase  
sections of a 2D data set independent of other sections.  That might  
solve your problem with already obtained data.  It seems like this  
would be easy to do with minor adjustments, and something I've wished  
for on occasion in the past.  I'll mention this to the Mestre-C  
people, who are busy about to make available a big new release that  
is Mac, PC and Linux compatible; we may purchase a site license if  
they've improved the stack plotting.
----
I am not familier with Bruker sequences.
However, "unphasable [i.e., small phase error that could not be  
corrected]" contours in a "pure-absorptive gradient coherence  
selected experiment"  could mean the p-type [for example, 4:1 ratio  
of gradients] and the n-type [corresponding 4:-1 ratio gradients]  
selections are not exactly the same.   This might mean that the  
positive and an equivalent negative gradients do not produce the same  
G/cm.   You may want to superimpose a positive gradient profile on  
top of a negative gradient profile and see how well they match.   In  
an ideal situation, they should superimpose.
----
I am aware of the phasing issue. One can usually phase the majority  
of peaks but some are out of phase. There is no processing procedure,  
that I am aware of, that would solve this. With a large range of  
intensities the problem always becomes more apparent, because you can  
no longer cut off high enough not to see the problem.
I think some of the sequences with adiabatic pulses do give cleaner  
spectra but I do not have real comparison data it is just my impression.
----
The main problem with the phase is the non ideal and offsett  
dependent 180°
pulse on 13C. You must take a pulse program with a "sp", that is  
applying an
adiabatic (frequency swept) chirp pulse, otherwise you will not  
succeed. See
experimental description in Exp. 12.10 of "200 Basic NMR experiments"
----
> ----- Original Message -----
> From: John Ralph
> To: ammrl_at_chemnmr.colorado.edu
> Sent: Friday, November 24, 2006 12:44 AM
> Subject: AMMRL: Contour peak phasing issues in hsqc spectra
>
> We have been having a problem for a while and, especially now when  
> we are trying to apply chemometrics methods to 2D NMR data, it is  
> becoming a real issue.
>
> It has to do with contour peak phases in 2D HSQC spectra.
>
> It seems we can never phase every peak in the spectrum -- if you  
> phase some, others are necessarily out of phase in a distinctly non- 
> linear way, particularly the intense contours.
>
> This is NOT an issue with a single spectrometer -- it is something  
> to do either with the types of sample and/or the parameters, and/ 
> or ?? The same "problems" are on 500-750 cryo machines in two  
> countries (and also on non-cryo at 360 MHz). We're using the ea si  
> hsqc expt (Bruker standard -- sorry, forgot the new name for a bit,  
> but used to be invietgpsi).
>
> The samples have a range of component intensities, are fast  
> relaxing (proton fid is dead in about 100 ms). We have tried the  
> obvious things (is relaxation delay too short -- no, 1s for  
> something that relaxes in 100 ms, AQ is 200 ms), turing p28 off or  
> not, and various other things. But we have not yet tried the  
> Adiabatic variants that may help if it is a J-mismatch problem  
> (with 180s).
>
> Anyone know how we might be able to solve this? (after acquisition  
> (by processing) would be great (e.g. is it that the early part of  
> the FID is messed up, but with digital we don't know how to  
> manipulate that the way we could in the old days), but before would  
> also help tremendously for the future). Any suggestions? Does  
> anyone need to see a dataset?
>
> Cheers and thanks in anticipation.
>
> John
------------------------------------------------------------------------ 
-----------------------------------------------------
John Ralph
US Dairy Forage Research Center, USDA-ARS        and Dept. of  
Biological Systems Engineering
1925 Linden Drive West                                       Univ. of  
Wisconsin, Madison
Madison, WI 53706-1108
(608) 890-0071  FAX (608) 890-0076
E-mail: jralph_at_wisc.edu
http://www.dfrc.ars.usda.gov
**Let me know if you want my vCard!**
Full-text Publications available directly from:
http://www.dfrc.ars.usda.gov/DFRCWebPDFs/pdfIndex.html
------------------------------------------------------------------------ 
-----------------------------------------------------
Received on Mon Nov 27 2006 - 10:13:21 MST

This archive was generated by hypermail 2.4.0 : Sun Jun 11 2023 - 15:14:25 MST