A rational method is presented for the selection of substituents to be introduced into a benzenoid ring system of a biologically active compound in order to a defined physicochemical parameter space. The method, which may be readily programmed for use on a computer, relies on maintaining a minimum distance between compounds in the multidimensional physicochemical parameter space. The series of compounds produced will then have a well-spread set of minimally correlated physicochemical parameter values and could thus be used for the reliable correlation of the variation in the biological activities of the members of the series with changes in these physicochemical parameters. Some examples of the use of the present method under various conditions are given, and it is compared with alternatives in the literature.

I A small group of arylamidinoureas for which the predicted physicochemical properties were widely spaced and uncorrelated was selected for study. 2 The antimalarial activity of each compound was measured against three species of plasmodium. Three corresponding regression equations were calculated relating the potency of each compound to its predicted physicochemical properties. 3 The potency of further test compounds was forecast for each species, by using the corresponding regression equation. These forecasts were compared with the measured activity of the test compounds, and were better than would be expected by chance. 4 It is suggested that only a few compounds in a series need be studied initially in order to derive regression equations with worthwhile predictive properties, if the compounds are selected according to appropriate criteria.

No abstract