Linearity of Pesticide Residue Test

[catlover]

Hi everyone,

I would like to ask for opinion on how to improve linearity on pesticide residue test via LCMSMS.
My target is to maintain linearity with R^2>0.995.
However, even I prepared a flesh standard, the linearity cannot maintain well.
What should I do to stabilize the signal and give good result?

The working range is 10 ppb to 200 ppb.
There are 34 MRM transitions in a single run.
(the time interval of each transition is 1 mins with RT as central time)

I tried to cut the number of transitions but the result does not improve at all.
The peak shape is ok, but I tried a back flush before, would this lower the performance of LC?

[Don_Hilton]

How does peak shape look at each end of the calibration range?

[catlover]

The peak looks symmetric at higher concentration, but in lower concentration say 10 ppb, 20 ppb, it is a bit asymmetric (not tailing but a bit like fronting).

In fact, I am doing internal calibration, and somehow the linearity is reduced if using internal calibration.
Should I increase the concentration of internal standard?

[Don_Hilton]

If your internal standard give a peak size in the middle of the range of your analytes in the calibration curve, you should be good.

Tell more about the analysisi - if you can, what the analytes are, at least compound classes. Tell how you made the standards serial dilution? All diluted directly down from a single solution?

How many data points across the peak? Do you switch between one set of transitions to another during the run or do you use the full set the whole way through? And if you have timed events switching sets of transitions, how close is the switch to the starts and ends of peaks?

[catlover]

If your internal standard give a peak size in the middle of the range of your analytes in the calibration curve, you should be good.

Tell more about the analysisi - if you can, what the analytes are, at least compound classes. Tell how you made the standards serial dilution? All diluted directly down from a single solution?

How many data points across the peak? Do you switch between one set of transitions to another during the run or do you use the full set the whole way through? And if you have timed events switching sets of transitions, how close is the switch to the starts and ends of peaks?

Thank you for your response.
Sorry for forgetting to tell you that I am doing with organophosphorus pesticide.
And all calibration solutions came from the same stock solution.
I make 10 ppb, 20 ppb, 50 ppb, 100ppb, 150 ppb and 200 ppb as conentration for calibration.
The data points are keep more than 12 per peaks.
And I cut the transitions into several time interval, but as I am using a 50 mm column, there are some overlapping between transitions.
I am sorry that I do not understand what is time events switch of transitions? Does it mean the ions scan rate?

I tried to cut the number of transition and isolated transition of internal standard from other transition.
Half of the result give better R^2 (at least around 0.99)
And the other sets seems contain signal saturation problem.
The curve of first 3 points are linear, and last 3 points are also linear (but lower slope), but with combining them into 6 points calibration, it gives a nonlinear curve.
I would tried to reduce either the capillary voltage or collision voltage of each transitions to see if it helps. (coz I do not want to only have working range from 10ppb to 50 ppb )

[Peter Apps]

There are two possible reasons for a poor fit to a straight line - either the response vs amount is truly non-linear, or the response for a given amount is not repeatable.

What is the repeatability for 5 replicates at the low end of the calibration ?

Peter

[catlover]

There are two possible reasons for a poor fit to a straight line - either the response vs amount is truly non-linear, or the response for a given amount is not repeatable.

What is the repeatability for 5 replicates at the low end of the calibration ?

Peter

I don't try the repeatability yet.
In the transitions, 4 of them gives 0.99 in every run while all the other seem have signal saturated. Thus I think the repeatability of low end should be ok.

[Peter Apps]

There are two possible reasons for a poor fit to a straight line - either the response vs amount is truly non-linear, or the response for a given amount is not repeatable.

What is the repeatability for 5 replicates at the low end of the calibration ?

Peter

I don't try the repeatability yet.
In the transitions, 4 of them gives 0.99 in every run while all the other seem have signal saturated. Thus I think the repeatability of low end should be ok.

To quote a post somewhere on the forum; "don't think, measure !" You will have to establish repeatability anyway in order to validate the method, so you may as well do it now when it can also contribute to troubleshooting the linearity problem.

Peter

[catlover]

Although I do not start the formal repeatability test (from sample preparation to analysis), I prepared 3 fresh calibration solutions of 10 ppb to check the stability of signal. The respond (peak area) were nearly the same, and so I started my work on calibration but there is now problem of non-linear curve.

And as I am using LCMSMS, and with other successful transitions, I am sure that the respond should be linearly corresponded to concentration.

[Peter Apps]

Although I do not start the formal repeatability test (from sample preparation to analysis), I prepared 3 fresh calibration solutions of 10 ppb to check the stability of signal. The respond (peak area) were nearly the same, and so I started my work on calibration but there is now problem of non-linear curve.

And as I am using LCMSMS, and with other successful transitions, I am sure that the respond should be linearly corresponded to concentration.

Indeed, it "should" be - but if it was, you would not be asking for solutions to a non-linearity problem.

"nearly the same" is not the result of a measurement. What was the relative standard deviation of the three replicates ?.

Peter

[Don_Hilton]

First - as Peter reccomends. Replicate injections will tell you a lot. They may even change the conclusion about linearity.

Next, Let me clarify the question on the timing of transitions.

On some instruments you can set up to acquire the transition from mass A to mass B from one time in the LC run to another in the run, call it (T1 to T2) - to acquire data for a compound ---and the transition for some other transition, call it mass C to mass D across some other time range in the LC run (T3 to T4) to acquire data for a second compound. If the peaks partially coelute, setting time windows for the two transitions to match the start and end times for the compounds can cause a problem if the number of transitions being acquired changes in the middle of a peak. (The order of events along the tromatogram would be T1, T3, T2 ,T4) If the apex of a peak shifts as a function of analyte concentration (or just run to run variation), things have the potential to get messy because the percent of total dwell time changes at T3 and T2.

If acquisition of data for all transistions are turned on or off well outside of any peaks in the chromatogram - there will be no problem from this. And I've listened to an instrument manufacturer claim that thieir instrument will allow you to make changes in transitions acquired and automatically takes care of the issue of changes in how many transitions are being acquired through a peak.

[James_Ball]

There are two possible reasons for a poor fit to a straight line - either the response vs amount is truly non-linear, or the response for a given amount is not repeatable.

What is the repeatability for 5 replicates at the low end of the calibration ?

Peter

I don't try the repeatability yet.
In the transitions, 4 of them gives 0.99 in every run while all the other seem have signal saturated. Thus I think the repeatability of low end should be ok.

Once the signal saturates the detector, you will have no useable information for quantitation. If there are compounds that saturate then those must work with a lower concentration calibration curve or the entire sensitivity must be lowered. It may be that you need to calibrate those compounds in a range of 1ppb-50ppb or 0.1ppb-25ppb or something similar while the others remain with a high calibrator of 200ppb. Mixtures of standards that have vastly different sensitivity are sometimes difficult to find the right combination of concentrations so that you can start from one mix to make the entire curve and keep them all in the range of the instrument.