ORAL PRES.: Chemometrics on metabolite profiles or fingerprints.

Worldwide, herbs are used for preventive and therapeutic goals. Therefore, identification and quality control of these products of natural origin is required. Determination of some of the active compounds does not always allow assessing their total intrinsic quality. Since 1991 the World Health Organization accepts fingerprint chromatography as identification and quality evaluation technique for medicinal herbs. In fingerprint development, the goal is to create general conditions to maximize the peak capacity within an acceptable analysis time. A fingerprint is an analytical pattern of an extract showing some common pharmacologically active and/or chemical characteristic features, i.e. showing the metabolic profile of the sample. It can be measured by a separation or a spectroscopic technique. Similarly fingerprints or metabolite profiles can also be developed for extracts from animal or human origin.
A fingerprint can be developed for a number of reasons: identification, classification or calibration purposes. Identification is to confirm that a sample is originating from the herb expected and to exclude that it is another, i.e. to attain a better quality control of the herbs. Classification can be performed to classify samples according to, for instance, their origin or to distinguish healthy from sick patients. For instance, one may be interested in the geographic origin of samples or to distinguish between natural and synthetic compounds, e.g. vanillin from herbal, synthetic or microbiologic origin. Such evaluation is most often done by a principal component analysis, occasionally by a cluster analysis. However, it might also be by building a classification model. A multivariate calibration can be performed when the extract also can be characterized by an activity, e.g. an antioxidant, a cytotoxic or an anti-inflammatory activity. The activity then can be modelled as a function of the fingerprints. The most commonly used modelling techniques are stepwise multivariate regression, principal component regression and partial least squares. The goal of the modelling can be either to build models that are able to predict the activity for future samples based on the metabolite profile (e.g. the antioxidant activity from green tea fingerprints) or to identify the main compounds/peaks in a fingerprint chromatogram, responsible for a given activity. In the presentation the different applications will be considered.