Pattern Recognition Applications in Chemistry and Pharmacology. III. Clustering of Substituted Glycolic Acid Esters (Glycolates).

摘要

A computerized pseudoclustering algorithm (SIMP) was written and applied to 12 chemical structure descriptors (features) which, on the basis of expert opinion, are required for substituted glycolic acid esters (glycolates) to have some type of physiological activity. The resulting clusters were then input to an algorithm (DISP) written to determine histograms of safety ratios (LP50/MED50) at specified levels and the overlap or confusion probability(Cp) between the histograms of the clusters and the general glycolate data base. Employing the results obtained from the SIMP and DISP programs a decision rule was devised which classified 421 glycolates at one of two levels of activity. The misclassification rate was 21% as compared to an expected rate of 39% based on random sampling. Similar results were obtained with respect to discriminating glycolates at one of two toxicity levels. The methodology appears general and applicable to any chemopharmacological structure-activity problem. (Author)