Why no noise in LC/MS application?

[Hanna Mint]

I have developed a method for a carbohydrate using ESI LC/MS. The compound do not contain any functional groups that easily (de-)protonate so I have used the sodium adduct for its detection. Nor does the compound fragment into ions that can be detected, so I have used SIR for quantitation instead of MRM. My problem is the following: When running sea water samples the matrix effects gets high, which not is strange, but in the same time the noise levels out and cannot be measured. In standard runs the noise looks completely normal. The peaks in the Sea water samples (that has correct retention times compared with standards) have areas that are about 10-1000 times higher compared with the area for a standard peak with an area that equals LOQ when no matrix is present. The method seems to work OK, the problem is that I cannot calculate a S/N ratio in the samples. (I have enlarged the noise with the software, but it still becomes nothing but a straight line). I use a labelled internal standard. The compound is chlorinated, so I aquire the two first peaks in the spectra, all in all three SIR transitions are monited in the same run. I guess cross-talk has nothing to do with this?

[sam.pedraglio]

I have a question to start with my guess: Have you an idea of the levels of your analyte in real sea water sample? I mean, if you consider the high matrix effect you've mentioned, is there a correlation between the peak you observe and the possible concentration?
If so, the noise absence could be directly related with the matrix effect that causes a strong ion suppression.

[Hanna Mint]

I have a question to start with my guess: Have you an idea of the levels of your analyte in real sea water sample? I mean, if you consider the high matrix effect you've mentioned, is there a correlation between the peak you observe and the possible concentration?
If so, the noise absence could be directly related with the matrix effect that causes a strong ion suppression.

Thank you for the fast answer! When quantifying the peaks (relative the labeled standard), I get results that well fits the results of others. I would like to ask if you know if this with no noise due to high matrix effects is mentioned anywhere in the literature? It would be interesting to read a journal article discussing this problem. Thanks again.

[sam.pedraglio]

I think you'd have no results searching for this issue in literature, my guess is strictly related to your high ion-suppression.
Another idea: are you sure you don't have discharge problem into your source? If you have a very salty sample, and with sea water I guess it is, you could generate discharge and then no signal.

[Hanna Mint]

Thank you very much your help. The samples originates from the Baltics where salinity is much lower than in southern of Europe and they have also been washed in an SPE-step. It must be that the same masses as my analyte exist in the matrix and constantly reaches the detector that is the cause of this problem. It is a bit annoying, since the method seems to work in every other aspect. Thanks again

[sam.pedraglio]

It must be that the same masses as my analyte exist in the matrix and constantly reaches the detector

in this case you must have noise and, eventually, a low S/N if you have low analyte levels in your samples.

the detector that is the cause of this problem

this should give low, or absent, signal in every conditions.

[Hanna Mint]

It is true that I have some broad "peaks" in my chromatograms. If these are noise, than I must define LOQ as 2 times S/N (if I want to stick with this method). Areas in quantified samples are at least 10 times larger than the area for a peak that is 10 times S/N in a standard run when no matrix is present. What is your opinion? Would you say this low value for LOQ definition is defendable?

[Ron]

What is the threshold set at? If you set the threshold too high you can easily lose the noise if there is ionization suppression.

[Hanna Mint]

Great! This seems to do the trick! When playing around with the threshold and smoothing settings the peaks in samples are at least 5 times S/N, after a quick try. Thank you both!

[lmh]

this is a nice example of why the S/N ratio approach to LOQ carries significant dangers. It can easily become completely meaningless. I much prefer methods based on the standard deviation of low calibration points.

[Hanna Mint]

this is a nice example of why the S/N ratio approach to LOQ carries significant dangers. It can easily become completely meaningless. I much prefer methods based on the standard deviation of low calibration points.

I do agree. Still: Quantifying low calibration points when no noise is visible is also a bit scary.