COMPUTER-ASSISTED ANALYSIS OF QUANTATIVE STRUCTURE-ACTIVITY RELATIONSHIPS OF DRUGS

A coding hypergeometric structure was devised for a set of 80 drugs,which belong in six pharmacological categories,e.g.analgesics anticholinergics,antidep-ressants, antihistamines,tranquillizers and antiparkinsonain agents,then super-Positioning of structure carried out one by one. The indexes of first-order valent molecular connectivity, calculated as the sum of the contributions of the bonds joining a given position were obtained at assigned positions 1-8 by topological index. Thus, the data matrix of molecular feature of each drug was generated. Principal component analysis was so performed on the 80 drugs with 8 decriptors for each drug that above high dimensional data matrix was reduced to four dimensions in terms of non-iterative reflection according to the expected percentage of explained variance. The calculated four—dimensional factor co-ordinates of molecules were also used to classify each molecule of 80 drugs as one of the six pharmacological categories according to Bayes' stepwise discriminant analysis. Sixty nine drugs were correctly disscriminated and this led to a classification correct of 86.25%.