RESUMO
Amphetamine samples were analysed by gas chromatography-mass spectrometry (GC-MS), and the peak areas of 33 target compounds were transformed by applying various pretreatment techniques. The objective was to optimise the ability of a number of distance metrics to establish links between samples of amphetamine originating from the same batch (henceforth referred to as linked distances). Furthermore, partial least squares discriminant analysis (PLS-DA) was used to evaluate the effects of various pretreatment methods on separation of amphetamine batches synthesised by the Leuckart reaction, reductive amination of benzyl methyl ketone, and the nitrostyrene route. The most efficient way to pretreat GC-MS data varied for the different distance metrics, although best results were obtained when data were normalised to the sum of peak areas, and either the fourth root or a logarithm was applied to the normalised data. When pretreating normalised data by fourth root transformation, Pearson correlation was the distance metric that was most successful at finding linked samples. Normalisation and the use of fourth root also represented the best method of pretreating data when employing PLS-DA to separate samples synthesised by different routes. To achieve a faster and more user-friendly procedure for evaluating chromatograms, experiments were performed in which the number of target compounds used to compare samples was reduced. The effect of each compound that was removed was studied by applying PLS-DA and by using Pearson correlation to calculate linked distances as well as unlinked distances (between samples from different batches of amphetamine). Considering both links between samples from the same batch and separation of samples synthesised by different routes, the best results were obtained with the data set comprising 26 compounds. Finally, it was found that the profiling method developed in this work was superior to an existing technique with respect to separating linked and unlinked distances.
