An interesting data processing question

[Travisdog2]

Hi,

I was asked the following (rewritten so that it makes sense) by a student and was a little bit stumped on the best way to answer. Your thoughts would be most welcome!

“SIM or MRM methods tend to utilise a single defined quant ion/transition and one or more qualifier ions/transitions for each analyte. In most instances the quant ion is used to define integrated peak area (response) and the qual ions are only there to provide qualification that the peak is substance X; they don’t themselves normally contribute to the peak integration.

The software that we use for analysis allows response to be derived from the quant ion only or alternatively the quant ion plus the qualifier ions. The latter obviously provides greater absolute response values, so why do most analyses seemingly define peak response based upon the qualifier ion only and what are the issues with using all ions?”

My initial thoughts are that responses based upon a quant ion only are likely to have less variation than those derived from all ions, but this would depend fairly heavily on the nature of the qual ions. I did play around with a few batches of data to prove/disprove the point and there seems to be some truth in this idea, but RSD gains by using a single quant ion v using all ions seemed to be minimal although the data was not low level. At the same time, increases in integrated peak area (in the case of some analytes with more abundant qual ions) were significant when all ions were used to integrate. I also considered SNR as a possible reason, and although the baseline of low-level quant ions is noisy, calculated SNR calculated (Peak to Peak or RMS) didn’t seem to change whether using all ions or a single quant ion.

Any thoughts on this?

Kind Regards

TD2

[H.Thomas]

Hi Travis,

as you mentioned, usually the qualifier will have a worse S/N and by including it into the integration, you will at best gain nothing or even lose S/N, thereby increasing your LOD. The detection limit is not about absolute peak areas or heights but S/N.
If you are working with low concentrations and your qualifier(s) give good signal(s), it might be worth a try to include them. I see no reason why this shouldn't be done if it improves your data.
Regards, Hartmut

[lmh]

I was going to say the same thing, but had given myself some time to think about it. The qualifier will usually be at lower abundance, so it's probably got a lower S/N ratio, and adding a signal with a worse S/N ratio to one with a better S/N ratio is not going to improve it.

But actually, in real life it probably makes next to no difference. It is highly unlikely that your qualifier will have a terrible S/N ratio, because if it did, it wouldn't work as a qualifier: it would give you out-of-range peak-areas too often, and you'd have too many incorrect rejections of the peak. In fact, by definition, the "noise" introduced by including the qualifier can't go outside certain bounds, because if the qualifier deviates too far from the expected value, the peak would be rejected. So in real life you're summing a good value (with a good S/N ratio) with a fairly good value (with a fairly good S/N ratio) and you're likely to end up with larger number and a pretty good S/N ratio!

On the other hand, questions like this provoke thought, and challenge how we've always done things. The whole idea of looking at a handful of masses stems from physical limitations. In a triple quad, we only physically measure a few transitions, and in a scanning instrument, in former times, we lacked the data-handling sophistication to do more than extract an ion chromatogram. If analysts were mathematicians, and if computers had been more powerful in the times when analytical software and methodology was getting established, maybe we'd be doing something more imaginative, like using some sort of PLS or Principal component analysis of the standard to establish a component (weighted vector of multipliers by which we multiply each mass) as the best measure of amount of compound. I am not a statistician, but this also combines nicely with purity analysis (i.e. if you do it using PDA data in a UV-based system, the unexplained variance during the peak is the impurity). It also very neatly allows full deconvolution of spectrally-different, partially-coeluting compounds (I think Shimadzu might have actually implemented this in their top-of-the-range PDA software, but Shimadzu don't seem to be good at telling us when they've innovated). Student questions are always good for us.

[Travisdog2]

Thank you both for some excellent and insightful answers...

I have spent much of the day playing with some data and it seems to confirm that it doesn't make that much difference whether the qual ions are included in the reported response or not. Low abundance qual ions obviously vary more than high abundance ones, but their contribution to the response is low (and they end up being flagged if they fall too far from expected values invalidating the result) whereas high abundance ions vary a similar amount to the quant ion so ultimately, it doesn't make all that much difference. In the examples I looked at RSDs varied between the two methods by small fractions of a percent.

[H.Thomas]

One more thought about where it would be useful to sum up the responses. Say we have a halogenated molecule, that gives us the parent ion with the loweset monisotopic mass, as well as the isotopic ions 2 and 4 amu higher, all at a comparable intensity. Say e.g. we want to analyse Bromoxynil in negative mode, so we have parent ions of 274, 276 and 278 amu with intensities of roughly 5050. If you use the transitions of all ions summed up, you would gain about 4x sensitivity, compared to only the lowest molecular mass ion.

[bcd_GCLCMSMS]

"Quantitative" and "Qualitative" ions are just labels to indicate purpose. Traditionally, the ion with the largest response is the used as the "quantitative" ion, but in cases such as: common isobaric interferences, poor precision (RSD), or poor signal-to-noise another ion may be better. Poor precision and poor signal-to-noise can happen to signals with large responses too; therefore, the traditional rule to selecting ions is somewhat flawed.

Summing of isotopic peaks is possible, especially for chlorine and bromine containing compounds, but my guess is that it rarely raises the signal-to-noise or RSD. Selectively summing an additional peak if it meets criteria would be a no go in my opinion. The method can't sum different numbers of peaks on different samples. A useful exercise would be to calculate the total analytical error of each data analysis method and select the best.

[James_Ball]

One more thought about where it would be useful to sum up the responses. Say we have a halogenated molecule, that gives us the parent ion with the loweset monisotopic mass, as well as the isotopic ions 2 and 4 amu higher, all at a comparable intensity. Say e.g. we want to analyse Bromoxynil in negative mode, so we have parent ions of 274, 276 and 278 amu with intensities of roughly 5050. If you use the transitions of all ions summed up, you would gain about 4x sensitivity, compared to only the lowest molecular mass ion.

Even with this approach you are still limited to response equal to the smallest detectable amount for the worst performing of the three masses. If 274 goes to zero abundance then you can't reliably quantify on a calibration below that point(actually above that point as you need some abundance to make it valid) otherwise you will lose the linearity of the calibration. If the single best ion gives a response of 1000 when the worst falls out and the combination of the three gives 2000 response, the higher response really doesn't mean anything as far as detectability increase, it just gives you a higher floor, similar to having higher baseline noise or not.

[lmh]

chlorine and bromine is an excellent example. If you summed the isotope peaks you'd get a bigger signal. But if you have a near-coeluting alternative that differs by one double bond, you might end up with a method that detects that too. So it's a balance between sensitivity and selectivity.

But look what happens if we add the chlorine isotope peak of a singly-chlorinated ion. We get 133% of the original signal, but we've also got extra noise, so we don't get so much improvement in the S/N ratio. Against this, we now have a second peak one third the size of the first, which we're potentially using as a qualifier. The moment it becomes too small to quantify reliably compared to the quan ion, the overall peak gets rejected, and since it's one third of the size of the main peak, this will probably happen at one third of the concentration that we'd start to worry about the main peak. So basically we've made the limits three times worse, for the sake of a 33% increase in signal, and less than 33% increase in S/N.

You need to sum peaks that are quite close in overall intensity for the process to gain anything.