Clustering analysis of keishibukuryogan formulas by use of self-organizing maps.

Kampo medicines, traditional herbal medicines in Japan, are comprised of multiple botanical raw materials that contain a number of pharmacologically active substances. Conventionally, the quality control of kampo medicines has been performed by analyzing the contents of two or three representative components. However, it is not sufficient to check quality only with a limited number of specific components. We performed HPLC of 287 lots of keishibukuryogan formulas, calculated the areas of 11 components on chromatograms as feature values and made a cluster analysis using self-organizing maps (SOMs). We verified the precision (repeatability and intermediate precision) of clustering results when using the same samples and successfully established an clustering method using SOMs that is capable of precisely clustering differences in HPLC-fingerprints among pharmaceutical manufacturers, differences in HPLC-fingerprints due to the combination ratios of botanical raw materials, and differences in HPLC-fingerprints due to changes in component contents caused by time-course deterioration. Consequently, we could confirm that this method is useful for controlling the quality of multiple component drugs and analyzing quality differences.