When it was released at the start of 2013, Progenesis CoMet v2.0 introduced adduct deconvolution to its metabolomics analysis workflow, prior to compound identification. But what is deconvolution and why is it so important?
What is deconvolution and why is it important?
Deconvolution in Progenesis CoMet is the process of grouping different adduct forms of the same compound, for the purposes of compound quantification and identification.
- For quantification, we’re typically only interested in the overall abundance of each compound in our samples, not in the separate abundances of the different adduct forms. By deconvoluting and looking at the compound as a whole, we also avoid misleading fold differences that can be produced by variation in ionisation between runs e.g. the ratio of protonated to doubly-protonated adduct forms may vary while the overall compound abundance remains the same.
- For identification, if we find more than one adduct form, we can calculate the compound’s neutral mass, immediately taking us a step closer to identification before we even perform database searches. Indeed, by identifying compounds based on their neutral mass (where possible), there’s none of the ambiguity that could come from different adduct forms having different putative IDs.
The overall result is greater confidence in both your measurements and identifications.
Deconvolution, by example
The deconvolution process itself is actually very simple. Before analysis, we specify the adduct forms (e.g. protonated) that we expect to find in our samples. Then, after peak picking, Progenesis compares each detected ion with each of its co-eluting ions; if their mass difference matches the difference between two adduct masses, there’s a good chance they’re adducted forms of the same compound.
As an example, let’s consider an experiment with only 2 adduct forms: protonated (M+H) and sodiated (M+Na). For simplicity in this example, I’ve chosen adducts that are singly charged, as it makes it easier to relate to the ion map (an ion’s m/z value will be equal to its molecular mass).
After peak picking, we can see a detected ion on our ion map at a retention time of 5.20 minutes and an m/z value of 437.1932. We know it could be either the M+H or the M+Na form of a given metabolite. For the moment, let’s call it M+X.
- If our ion is the M+H form, then the M+Na form, if present, should be at an m/z of 459.1751. This is calculated by taking the mass of M+X (437.1932) then subtracting the mass of H+ (1.0073) and adding the mass of Na+ (22.9892).
- If our ion is the M+Na form, the calculation is simply reversed; we take the mass of M+X then subtract the mass of Na+ and add the mass of H+, giving 415.2113.
So, we now know 2 places on the ion map where we might find other ions of the same compound. Looking at the ion map in the area around M+X and the potential adduct locations, we see:
Our example ion, M+X, seen among a number of co-eluting ions on an ion map. Its isotopic peak has an m/z of 437.1932; the positions of other potential ions of the same compound are marked at 415.2113 and 459.1751.
Looking at the above, we can see that there is a co-eluting ion with its monoisotopic peak at an m/z value of 415.2113. Great! This means that M+X must be M+Na, while the ion with the lower mass must be M+H:
The deconvoluted compound ions highlighted on the ion map, showing the adduct form of each ion
Using this information, Progenesis then groups these two ions as the same compound, assigning the compound ions their respective adduct forms. Not only that, but we’re able to calculate the compound’s neutral mass:
Neutral mass = (437.1932 – 22.9892)
Neutral mass = 414.2040
And that, in a nutshell, is the deconvolution process that Progenesis CoMet performs automatically, immediately after peak picking. Every ion and all possible adduct differences are considered, leaving us with a list of compounds that we can then identify.
Of course, for adducts with different charges, the calculations are slightly more complicated, and there are always tolerances to take into account for both m/z and RT values, but it’s still a relatively simple process. And it’s made all the more simple for being fully automatic.