Chemstation - MS Background Subtraction trick?

[Lab029]

Hi everyone! I have a tough question about the background subtraction feature in Chemstation (v E.01) on an Agilent GC-MS. I have two species that are co-elute right on top of each other (the apexes are only a hair off). I have a standard of my unwanted co-eluting species and that has been ran in a different data file of course. It would be ever so handy if I could run a background subtraction on my sample spectrum by using the spectrum from my standard run so that I can get a nice, clean spectrum of my target species that is buried under the unwanted species.

I am trying to get a full mass spectrum of my target species for publication so I am being pretty particular about getting the full spectrum at the apex of the TIC. When I run the background subtraction off the shoulders I can get my unwanted species out but I am still grabbing some of my target species MS enough that it alters the spectrum quite a bit so the M/Z intensities are all wrong.

I seems impossible but I was curious if anyone has found a trick to do it....maybe a macro exists? Can some some other software run a background subtract function from two different mass spectra? My sample was ran over a year ago and I no longer have the source material so I can't just re-tool the method for better separation so I need to find a more creative solution. In the end, I suppose I can copy the spectrum and paste it somewhere to manually erase the M/Z belonging to the other species but I would like a more automated / consistent function. Anyway, thanks for the help!!

[shfo]

You would have to export the spectra from Chemstation first, but I think (I really haven't used it much) mmass has some tools for cleaning up and background correcting spectra.

[James_Ball]

One easy trick would be to inject the interfering standard a minute after you inject your test sample, then the peak you want to subtract would elute after your peak of interes. Then you just have to do a normal subtraction to get the clean spectra.

If you can judge the concentration of the interfering peak you can also make a standard at or just above that concentration so that you get a complete subtraction of the unwanted spectra.

[Yama001]

You can manually subtract any spectral or chromatographic objects you like. If you turn on the variable watch (STACK command), you can see how the various spectra are stored.

For example - load your data file and grab the spectrum - it will be put into the X variable and displayed on screen (usually window #2). Then load file that has the peak you wish to subtract and grab that spectrum - it should and up in the X variable and the previous spectrum will roll down to the Y variable. Subtracting is as simple as "T=Y-X". You may then view the result with "DRAW 2, T".

The menu driven procedures use a similar scheme; they are easier to use as they will handle the syntax, but you can get a bit more flexibilty by learning to use the command line. Check out help for the various commands. You may also save any variable you like using the STORE and RECALL commands. The may be commands that will allow you to removed mass peaks manually as well, but I do not recall if those are still available.

The software will always label the graphics with some sort of identifier to cluse you in the the data has been manipulated in some manner, just like you would see with an automatically background subtracted run.

You cannot use this type of scheme to automatically put this stuff into the canned reports, but if you just need a cleaned up spectrum, this approach gives you alot of flexibility.

[lmh]

just subtracting won't do, because you need to subtract exactly the right amount. I like shfo's idea of trying some external software.
Mathematically what you're trying to do is a regression where you want the residuals.

Measured = Standard*x + Unknown

Measured, Standard, and Unknown are vectors, x a single number, and you want to do a best fit choosing x to minimise the total excess of Unknown, and then display Unknown. If you can find a friendly statistician/numerical person, and export the two spectra, it's not difficult for them!

[lmh]

Been thinking about this, and of course if your two compounds genuinely coelute to the extent that you don't have spectra containing different amounts of them (more different than random variation in spectra) then there are some circumstances in which it can be at best an approximation.

For example, imagine the extreme situation that your known spectrum contains only one mass; it's possible this mass is also present in the unknown spectrum, but there is no way to tell how much of that particular mass comes from the "known"... It's easy to see that this applies to any situation where the known spectrum consists of a subset of masses found in the unknown, and since you don't know the unknown, in theory you can't be 100% sure it hasn't happened.