PREDICTION OF PHARMACOLOGICAL CLASSIFICATION BY MEANS OF CHROMATOGRAPHIC PARAMETERS PROCESSED BY PRINCIPAL COMPONENT ANALYSIS

Based on linear free-energy relationships (LFER), it has been assumed that systematic information on the behavior of a series of xenobiotics in a number of appropriately designed physicochemical systems can be used to predict the differences in their biological activity. Computer ized methods of multivariate data processing, like principal component analysis (PCA), allow for extraction of such systematic information f rom large sets of diverse and mutually interrelated physicochemical pa rameters of drug analytes. High-performance liquid chromatography (HPL C) is a unique method that can readily produce a great amount of physi cochemical data on a large set of analytes. Modern HPLC methods allow for inclusion of biomolecules as the active components of a chromatogr aphic system. A group of 83 drugs of established pharmacological class ification were chromatographed in eight carefully designed HPLC system s. A matrix of 83 x 8 HPLC data was subjected to a statistical analysi s by PCA. A grouping (clustering) of drug analytes was obtained which was exclusively due to a systematic similarity of their behavior in th e HPLC systems studied. The obtained clustering of drugs was in accord ance with their pharmacological classification. Specific bioactivity f eatures emerged for several drugs which have been rationalized in view of literature reports. A hypothesis has been put forward that a multi variate analysis of HPLC parameters may help to segregate drugs and dr ug candidates according to their pharmacological properties and thus t o appropriately guide the testing and to limit the number of routine b iological assays. (C) 1997 Elsevier Science B.V.