Galvez and colleagues have demonstrated in several publications that the SIR descriptors can be used effectively in discriminant analysis.22'46'47 In addition, they have shown that screening a structure library found compounds that proved to be active.
A database with compounds in seven different pharmacological classes of activity was used for development of discriminant models of each activity class. The classes included analgesic, antiviral, bronchodilator, antifungal, hypolipidemic, hypoglycemic, and beta-blocking activity. Galvez developed separate discriminant models for each class by using molecular connectivity indices. Based on each model, activity was predicted for a list of structures as both a prediction and a subsequent experimental test. In some cases, compounds were also tested experimentally. Compounds predicted to be active were generally known from the literature to be active or tested in the laboratory.
For example, for antiviral activity, a library of over 12000 commercial compounds was screened. Seventeen compounds predicted to be active were selected to be tested. In an in vitro assay (herpes simplex-lvirus on cellular cultures) 12 compounds were found to significantly active, with activity similar to that of phoscarnet. These compounds included nitrofurantoine, 1-chloro-2,4-dinitrobenzene, 5-methylcytidine, 1,2,3-triazol-4,5-dicarboxylic acid, cordicepine, nebularine, and inosine. Galvez concluded that some of these compounds might be considered as lead candidates for new drugs. Similar results were reported for the antibacterials and analgesia.
Table 9 Similarity screening results for Albuterol, based on eight atom-type E-state descriptors for all eight atom types in the target structure, Albuterol
Was this article helpful?