Hello all,
The data shown here is a hypberbole to demonstrate a problem I've observed.
I was wondering if it was possible to change topspin's peak picking strategy from a "bottom up" approach to a "top down" approach. Attached is an example of the phenomena I've observed.
Setting MI to a very low value (in this case, 5e-5), the peaks are picked in a way such that peak #1 is the LEAST intense. Not a problem, since all the peak labels are is a label right?
Except that in cases where I adjust the PPRESOL to a higher integer. I was hoping to get rid of the shoulders of peaks. Instead, I'm observing an opposite effect where the shoulder is a lower intensity and is therefore picked first. Next, when I would expect the peak maximum to be picked, if it is within the 10 points that the PPRESOL defines, then the peak maximum is NOT picked at all.
What exactly is going on here, and how would you change the "direction" of this so that it picks the maximum intensity peak over the lower?
Hallo Joseph,
In my eyes topspin is not the ideal program to perform data analysis like peak picking. Do you have the possibility to use Mestre Nova? For 1D and 2Dspectra it is verry convenient. otherweise there are also free programs available like CcpNmr.
Many greetings
PPRESOL will not help. The only thing this does is not to list a peak that is within n points of a peak it found. And you are correct it will prefer the low level peaks.
The program should never pick shoulders unless they are really a peak. This means they are a local maximum. If it is a should that is not higher than the surrounding points it should not get picked. I will need to check if there is away to change the direction of peak picking.
Clemens Anklin , Thanks very much for your feedback on this. I find that the peak picking is the number one issue that I am having, and has the greatest variability in the results I'm getting downstream in my metabolomics work (logically so).
I've tried adjusting the PPRESOL to a very low number (usually between 1 and 3) which would pick all "possible" peaks, then filtering by maxima in a localized area, but am weary when thinking about how this might work for homonuclear 2D spectra which could have very close associated signals.
Joseph,
As an idea (note that I have no actual experience with this specific problem), perhaps it would work to do your peak picking in 2 (or more) stages. First, you set a relatively high MI and do the peak picking, then set it to a relatively low MI and execute it again. Perhaps it will then recognise that there is already a peak from the first stage, and not mark a peak in the second stage. This could easily be scripted as a macro, python, or au program.
- Badges
- Science topic
- Similar topics
- Physical Sciences
- Atomic, Molecular and Optical Physics
- Atoms