The relationship between the mass spectra of drugs and their biological activity—An application of artificial intelligence to chemistry

The mass spectra of a set of drugs consisting of sedatives and tranquilizers are analyzed and differentiated by pattern recognition methods, such as nonlinear mapping, K-nearest neighbors and Fisher discriminant. The mass spectral peaks are ranked, selected and weighed by the Fisher Ratio of each peak. The results indicate that it is possible to separate the two types of drugs and to predict the pharmacologic activity of “test” drugs with a high degree of accuracy. These general computer methods can be applied to other drugs or any data which requires separation into two classes.