Pharmacological classification of drugs based on neural network processingof molecular modeling data

The performance of artificial neural network (ANN) models in predicting pha rmacological classification of structurally diverse drugs based on their th eoretical chemical parameters was demonstrated. The classification coeffici ents for psychotropic agents, beta -adrenolytic drugs, histamine H-1 recept or antagonists and drugs binding to alpha -adrenoceptors were 100, 100, 95 and 86%, respectively. A set of easily accessible non-empirical molecular p arameters describing the structure of xenobiotics can provide information a llowing the prediction of some pharmacological properties of drugs and drug candidates employing ANN models. Since ANN analysis can help cluster as we ll as segregate drugs and drug candidates according to their known and expe cted pharmacological properties, the number of routine biological assays mi ght be reduced. The results presented here might be used to improve the eff iciency of high throughput screening programs for new drug hits by demonstr ating a promising procedure for diverse combinatorial library design and ev aluation.