117828230

528 M. Daszykowski et ai

To interpret differences among the groups of sagę samples, score and loading plots are interpreted simultaneously. The largest contribution to the first PC (PCI) is the peak of retention time, te, 11.77 min; the largest to the second PC (PC2) are the peaks of Ir 13.28 and 13.49 min. We have previ-ously shown that these peaks are those of camphene, limonene, and eucalyptol, respectively [40]. Samples from the first group have a relatively large peak at 11.77 min compared with samples from the other groups and, hence, these particular samples are relatively rich in camphene. Samples from the second group are characterized by medium amounts of camphene and very Iow amounts of limonene and eucalyptol. The unique properties of the samples from the third group can be explained by the loading values on PC2. For these particular samples, peaks with te 13.49 min (high inten-sity) and 11.77 and 13.28 min (both of Iow intensity) seem significant. These samples contain relatively smali amounts of camphene and limonene, but larger amounts of eucalyptol. Fig. 5c shows a selected time region for two superimposed chromatograms, those from samples 7 (dashed linę, S. cad-mica) and 14 (solid linę, S. nemorosa), for which Chemical differences were identified by use of PCA. Chemical differences between the two groups of fingerprints are mostly quantitative in naturę.

For the HPLC fingerprints, compression by means of PCA was quite ef-fective, with the first two PCs explaining over 72% of the total data vari-ance. The most interesting pattern of samples is revealed on the PC1-PC3 score plot (Fig. 5d). Two samples have unique properties, and are located far away from the main bulk of the data. These are the extracts of S. glutinosa (sample 13, with the large score value on PCI) and S. officinalis (sample 20, with a large score value on PC2). The uniqueness of S. glutinosa can mostly be explair»ed by the presence of a relatively high peak at te 18.75 min whereas for S. officinalis a relatively large peak elutes at te 11.92 min (Fig. 5e). Selected regions of these two chromatograms are presented in Fig. 5fi the differences between them are very distinctly evident.

Conclusions

Preprocessing of chromatographic fingerprints is a crucial step which can be regarded as compulsory before commencing comparative analysis by use of chemometric techniques. Depending on the type of chromatographic sig-nal and the problem scrutinized, different preprocessing of the chromatograms can prove necessary. When morę than one type of preprocessing is required, however, it is necessary to choose the sequence of different preprocessing methods carefully. In this paper, we used a four-step preprocess-