Processing and analysis of chromatographic data

Data pre-processing and analysis techniques are investigated for the analysis of one- and two-dimensional chromatographic data. Pre-processing, in particular alignment, is of paramount importance when employing multivariate chemometric methods as these techniques highlight variance, or changes between samples at corresponding variables (i.e. retention times).

Principal components analysis (PCA) was employed to evaluate the effectiveness of alignment. Two methods, correlation optimised warping and icoshift were compared for the alignment of high performance liquid chromatography (HPLC) metabolite data. PCA was then employed as an exploratory technique to investigate the influence of phosphite on the secondary metabolites associated with Lupinus angustifolius roots inoculated with the pathogen, Phytophthora cinnamomi.

In a second application, HPLC with acidic potassium permanganate chemiluminescence detection was evaluated for the analysis of Australian wines from different geographic origins and vintages. Linear discriminant analysis and quadratic discriminant analysis were used to classify red and white wines according to geographic origin. In the analysis of wine vintage, partial least squares and principal components regression were compared for the modelling of sample composition with wine age.

Finally, software was developed for quality control (QC) of flavours and fragrances using comprehensive two-dimensional gas chromatography (GC×GC). The software aims to automatically align and compare a sample chromatogram to a reference chromatogram. A simple method of partitioning the two-dimensional pattern space was employed to select reference control points. Corresponding control points in a sample chromatogram were identified using a triangle-pattern matching algorithm. The reference and sample control points were then used to calculate the translation, scaling and rotation operations for an affine transform, which is applied to the complete sample peak list in order to align reference and sample peaks. Comparison of reference and sample chromatograms was achieved through the use of fuzzy logic.

It is concluded that the pre-processing and chemometric methods investigated here are valuable tools for the analysis of chromatographic data. The developed GC×GC software was successfully employed to analyse real flavour samples for QC purposes.