It is difficult to understand precisely the physiological actions of herbs because they contain a complex array of constituent molecules. In the present study we used DNA microarray data for 12600 genes to examine the anti-proliferative activity of the herb Coptidis rhizoma and eight constituent molecules against eight human pancreatic cancer cell lines. We identified 27 genes showing strong correlation with the 50% inhibitory dose (ID50) of C. rhizoma after 72-h exposure. Hierarchical cluster analysis with correlation coefficients between expression levels of these 27 C. rhizoma-related genes and the ID50 of each constituent molecule classified these test molecules into two clusters, one consisting of C. rhizoma and berberine and the other consisting of the remaining seven molecules. Our results suggest that one molecule, berberine, can account for the majority of the anti-proliferative activity of C. rhizoma and that DNA microarray analyses can be used to improve our understanding of the actions of an intact herb.