Evolving Factor Analysis (EFA) is a promising multivariate technique for model-free computational resolution of overlapping chromatographic peaks [1]. The number of components is found by following the emergence and disappearance of the different "patterns" in the data. The intrinsicly ordered nature of chromatographic data is then used to calculate the elution profiles and pure spectra of the compounds in the mixture. The large amount of uniquely "patterned" information in a mass spectrum makes the combination of EFA and GC/MS ideal.
GC/MS/EFA offers advantages for the mathematical separation of components in an unknown mixture as it makes no assumptions about the number of components, their identities or their chromatographic behaviour. This is potentially useful in applications such as large-scale pollutant and drug screening where the sample throughput is a problem and accelerated GC/MS followed by automated deconvolution, identification and flagging of target compounds would be efficient as well as cost-effective.
Simulated and experimentally-based GC/MS data have been used to investigate the capabilities and limitations of EFA as applied to GC/MS [2]. Data has been acquired on a magnetic-sector instrument (VG16F), a quadrupole instrument (VG-Quattro) and a time-of-flight mass spectrometer (OA-TOF). The method has been rigorously tested: in the mass spectral dimension (set of n-alkanes with very similar spectral "patterns"); the time dimension (ultrafast narrow-bore capillary GC/OA-TOF - 10 compounds acquired in a 2s time window); and in its dynamic range. Noise has been shown to limit the ability of EFA to reveal directly the number of components in an unknown mixture. In cases where the number of components is already known however, the EFA method may be applied with remarkable success despite very noisy data.