Amdis: one peak, two components

[sesquiterpene]

Hello everybody,

I'm using Amdis to analyze some data.
In some cases (often when peaks are tailing etc.), the deconvolution process yields two components which are clearly the same compound (Example: https://drive.google.com/file/d/0B7Rtha ... sp=sharing).

My analysis definitions are already as inclusive as possible: low resolution, very low sensitivity and low shape requirements.

My assumption is that I can't make Amdis treat the two components as one (though I'd be happy to be corrected here).

But how would you deal with this situation if you wanted to quantify the compound?
In this case, the component's areas are similar and clearly non-negligible. Would you add them? Use an average?

Thanks!

[chemstation]

Did you also try changing "Component Width" ?

From the AMDIS help file:
Component width Default = 12

This is the width that a component is assumed to be and is expressed in terms of mass spectral scans. If you are analyzing data with very wide peaks, you may want to increase the default value, or if all of your data are very narrow, you may wish to decrease it slightly. In general, larger settings will slightly slow down the analysis, but the effect is small. In any case the analysis will automatically adjust for wider peaks to some degree

[sesquiterpene]

Did you also try changing "Component Width" ?[/i]

Thanks for the advice.
I haven't because it was set to the max (32) and I thought that the larger the value is, the more likely it is that Amdis considers a wide peak to be only one component.

I tried taking it down and saw that decreasing it to 6 partly solves the problem, but at the price of missing some components.

[chemstation]

You can now answer your initial question of whether you can add the areas together.

Do you want to quantify or are you doing principle component analysis (PCA) and using area ratios.

As quantification requires reference standards, calibration curves etc..., that the GCMS
software would do as default.

[sesquiterpene]

Do you want to quantify or are you doing principle component analysis (PCA) and using area ratios.

As quantification requires reference standards, calibration curves etc..., that the GCMS
software would do as default.

I'm not sure exactly which analyses will be done because I'm only in charge of the initial data analysis in this project. But does it really matter? In both cases I'd need some good and consistent read of the component size in area units. Does it make any difference if I need it for an estimation of absolute amounts or for comparison to other components (i.e. area ratios)?

Anyway, I think I figured out something. I found the option to show the component on the chromatogram. When graphically comparing the two components which are actually one, it is in most cases clear that they are not mutually exclusive - i.e. they seem to overlap strongly. This implies that simple addition of the areas would yield an overestimation of the component's size. In most cases the component based on a peak (rather than the TIC) continues further into overlapping components and has a more realistic shape. It also has bigger area than the other component. Thus, it seems that taking only the bigger read gives the best estimation for the component's size.

[Peter Apps]

I gave up on AMDIS a while ago, and I now use AnalyzerPro from Spectralworks.

It's not going to help much, but this is not really a data processing problem - the problem is with the chromatography. You have very wide, tailing peaks with hardly any signal:noise and what is probably a high and fluctuating background for whatever ion is shown in white - which does not rise and fall in synchrony with the other ions.

Peter

[sesquiterpene]

You have very wide, tailing peaks with hardly any signal:noise

Thanks, Peter.
The example that I gave was indeed extreme, but I've had it in the past with higher-quality runs. Anyway, since these aren't liquid samples, we won't be able to rerun them using different concentrations or oven programs, so I'll have to live with what I have.

[Peter Apps]

You have very wide, tailing peaks with hardly any signal:noise

Thanks, Peter.
The example that I gave was indeed extreme, but I've had it in the past with higher-quality runs. Anyway, since these aren't liquid samples, we won't be able to rerun them using different concentrations or oven programs, so I'll have to live with what I have.

I feared that might be the case.

I don't know if you can do it with AMDIS, but in AnalyzerPro you can eliminate certain ions from analysis. Something like m/z 60 which is common to a lot of compounds and contributes a lot if baseline noise actually does not help much to find peaks, so if you take it out the deconvolution can still find the peaks, but the wobbly baseline is no longer a problem.

Peter

[sesquiterpene]

I don't know if you can do it with AMDIS, but in AnalyzerPro you can eliminate certain ions from analysis.

Amdis allows you to specify up to 8 m/z values which should not be used as model peaks.

[Peter Apps]

I don't know if you can do it with AMDIS, but in AnalyzerPro you can eliminate certain ions from analysis.

Amdis allows you to specify up to 8 m/z values which should not be used as model peaks.

Omitting m/z 60 from the model has to be worth a shot.

Peter

[sesquiterpene]

Thanks a lot!
I did a very quick experiment, and indeed omitting some ions solves the problem: I manage to get only one peak. I am yet to explore whether this has any negative effects globally - i.e. when analyzing the entire chromatogram and/or other samples.

Yet from what I see so far, the calculated area is very similar to the higher value from the multiple components. This indicates that my strategy of allowing Amdis to identify the compound "twice" and then removing the lower-area ones is working reasonably well.

[Peter Apps]

Thanks a lot!
I did a very quick experiment, and indeed omitting some ions solves the problem: I manage to get only one peak.

Excellent !

If you want a reprint with some more detail on deconvolution strategies e-mail me at
peterjamesapps - at - gmail-.-com (take out the anti-bot dashes and spaces of course)

Peter