Cluster significance analysis contrasted with three other quantitative structure-activity relationship methods

Abstract

Cluster significance analysis (CSA), a new statistical method to analyze structure-activity relationships in graphically displayed data, is contrasted with linear discriminant analysis, SIMCA, and the method of "relative odds". The data sets evaluated are as follows: antibacterial lasalocid derivatives, antimalarial naphthoquinones, and carcinogenic polycyclic aromatic hydrocarbons. CSA gives results comparable to these other methods, involves fewer assumptions, can be more reliable, and in general is easier to understand.