Using water matrix for construction of calibration curves

[k.schmidt]

Hello,

I apologise if this topic was previously discussed here (I couldn't find anything relevant however).

I want to construct calibration curves. Let's say that the analysis concerns the SPME of volatiles in wine and I am building the calibration curves for some of the detected volatiles only as I couldn't identify all of the VOCs present in the sample.

In many papers researchers studying VOCs in wine (or beer) use water as a matrix to built the calibration curves. Why not wine? There are competitive adsorption and absorption interactions between the analytes on SPME fiber. There are more VOCs present in wine than in water model. Therefore these interactions might differ between the two matrix types. Could I use wine spiked with different levels of standards to built calibration curves? Then I'd subtract the peak areas of the analytes from the wine blank from the peak areas of the analytes from the spiked wine samples to build the calibration curves.

Or is water just simply agreed as a "standard matrix" for water-based samples as it surely makes the analysis easier? I suppose if back-calculated accuracy and precision from wine samples are demonstrated to be in the acceptable range, then it's fine to use water for construction of calibration curves.

Will be very grateful for any comments. Thank you.

Kamila

[Peter Apps]

Hi Kamila

Known addition to wine is certainly the most accurate approach. Failing that I would use aqueous ethanol with the same alcohol content as the wine.

Peter

[k.schmidt]

Thank you Peter.

[BMU_VMW]

For a shorter period of time it might work to use actual wine. But over a longer period there will be a lot of variation on the used matrix (other bottle, older, other year 'production', ...) .
Water, with EtOH added to it will be much more constant and thus easier to use.

[James_Ball]

Another thing to consider is that some regulating agencies such as FDA or EPA do not allow blank subtraction of results. Even though we know that standard addition methods work just fine.

If you could purge the wine of the volatiles you want to monitor so that you have the water, ethanol, and other non-volatile constituents but have the targets of interest removed, that would make the best calibration matrix you could use. The problem with that is the difficulty of removing the volatiles without altering the matrix.

[k.schmidt]

I have asked about wine as there are many publications available. However, I am actually working with the cell culture media as a sample. I have not found any paper regarding this matrix and method validation. Its main constituent is water but obviously not only.

I am afraid of using this matrix to build my calibration curves because of the variables pointed out by BMU_VMW. My research has nothing to do with FDA or EPA so hopefully I would go away with subtractions (with good reasoning behind it), however if it's not allowed...

I think I'll go for water...

[Peter Apps]

I have asked about wine as there are many publications available. However, I am actually working with the cell culture media as a sample. I have not found any paper regarding this matrix and method validation. Its main constituent is water but obviously not only.

I am afraid of using this matrix to build my calibration curves because of the variables pointed out by BMU_VMW. My research has nothing to do with FDA or EPA so hopefully I would go away with subtractions (with good reasoning behind it), however if it's not allowed...

I think I'll go for water...

Why, Oh why, didn't you tell us what you are doing in the first post ?? Why ask us about wine when you want to work with cell cultures ??

Water is a bad choice. If you are only interested in the media then you can use blank media and known additions, if you want to determine metabolites or depletion of substrates then you have no choice but to run a set of known additions to each sample, because the nature of the medium will vary from sample to sample.

Peter

[k.schmidt]

I'm sorry Peter. I've asked about wine because essentially it is the same principle as in my research so it's easier to explain. However, obviously wine is not the same matrix, my fault...

I am investigating volatiles in the cell culture media produced or consumed by the cells. This means that although at different concentrations than from media cultured with cells, the analysed volatiles are ALWAYS present in the blank media. That would mean that I have to do subtractions if I use media as a matrix for the method validation...

To complicate the matter one blank for subtractions wouldn't be enough. As you say each blank medium may vary from sample to sample so I would have to always run the control for the subtractions. Water would be just so much easier, but so much a different matrix...

Therefore, do analytical chemists run known additions with subtractions in such cases as this? Because if it's not approved by the regulating agencies some would still go for water I suppose.

I'd like to do it properly (although it doesn't seem to be so straightforward as I imagined). It's just the matter that I don't know what is considered as 'good practice' in the real labs out there....

[Peter Apps]

For each sample you run it as is, and spiked with three (or more) levels of analyte. You then plot peak area vs amount added, and divide the intercept on the y-axis by the slope to give the amount in the sample.

Peter

[k.schmidt]

I understand that standard addition is the best method for building calibration curves in the case of my samples. The problem is that firstly, my instrument's response is not prefect so I need to use internal standard. Secondly, I also need to validate my method. I cannot determine limits of detection if I use cell culture media as a matrix.

I am sorry if "I'm thinking loudly here". I've never validated a method in my life before...

[Peter Apps]

You are not really building a calibration curve in the usual sense of a set of data that samples are compared against. You make a short "calibration" series for every sample by the 3 or more levels of analyte addition.

There is no point in determining limit of detection when your matrix contains the analyte(s) already in detectable quantities. What you need to establish is the smallest difference in analyte concentration that you can measure.

Peter

[James_Ball]

You are not really building a calibration curve in the usual sense of a set of data that samples are compared against. You make a short "calibration" series for every sample by the 3 or more levels of analyte addition.

There is no point in determining limit of detection when your matrix contains the analyte(s) already in detectable quantities. What you need to establish is the smallest difference in analyte concentration that you can measure.

Peter

If I am thinking correctly here, then to accomplish this he would need to run a media sample, then spike it with three different levels of analyte to establish the "calibration". Then to establish the "lowest limit of variance detectable"(I made up an new term there, like we need any more) he would then analyze the media with let's say, the middle spiking level and do 5-10 replicates, take the variance on those replicates and use that versus the "calibration" and the result would be the lowest limit of variance detectable.

[Peter Apps]

My approach would be from either of two directions.

Decide on how small a difference you need to detect in order to answer the research question you need answered - so if you are expecting the bugs to produce thousands of ppm of product you do not need ppb differentiation limits. The divide that by something that is both convincing and sensible - so if you expect 100 ppm, prove you can find 10 ppm.

Or

Spike some matrix at a known low level, look at the signal:noise and work out how much lower you can go. Go to that level and if it works run five (or more) replicates and show that rsd is less than 10%.

Peter

[k.schmidt]

Hi,

I am going to extend this topic. Sorry if this is a basic question but I have managed to confuse myself. How to calculate % recovery in standard addition method of calibration? I have some of the compound of interest present in my matrix from the start.

%Rec = (measured concentration / theoretical concentration) x 100%

Measured concentration: I've got my calibration curve. I can calculate the added concentration of the spike from the equation. Is it the "measured concentration"?.

Theoretical concentration: Is this the amount of the standard I have added?

What about the analyte's concentration present in the matrix from the start?

The amount (volume) of the standard added is negligible.

Thank you very much,

Kamila

[Peter Apps]

"Recovery" is irrelevant in standard addition methods.

Peter