I have been working in developing methods for chromatographic optimization (follow this link for more information). One interesting aspect of this methodology is the development of objective functions that are expressing correctly the degree of overlap in a chromatographic separation. This is particularly difficult in two-dimensional chromatography. This is mainly becasue of two reasons. First, when going from one-dimensional to two-dimensional chromatography, the concept of “peak vicinity” changes. Indeed, although in one-dimensional chromatography we can consider only two neighbours (obviously one behind and one ahead), in two-dimensional chromatography there is no limit to the number of neighbours that we can consider. That has some consequences on how do we measure overlap. Second, because the peaks are not Gaussian, we need a different metric for the overlapping degree of peaks. More specifically, the traditional valley-to-peak ratios will fail, since the “valley” is not necessarily in line with the two peaks.
In our research we cam across with an interesting solution for two-dimensional valley-to-peak ratio. The measurement takes into account not the “valley” between two maxima, but the saddle point between them (we could call this the saddle-point-to-peak ratio). In this piece of work, we were able to correlate the error in chromatographic peak area measurement with this resolution metric. We applied the resolution metric to GCxGC data, and it can be easily applied to HPLCxHPLC data. Application areas are broad (specially oil & gas, food and forensics). Refer to publication [22a] for more information.
This project was developed at the University of Amsterdam. Several people were involved. See authors of the publications for more details about authorship. The work of using cross-validation to assess the number of compounds was part of the MSc thesis of Sonja Peters [1p].
See my presentation in Dalian (GCxGC-2007).