The use of structure generators in predictive pharmacology and toxicology.

The analysis of pharmacological effects of chemical compounds often requires expensive of database searches. But in many cases, some of the most useful (or toxic) substances are overlooked because they are missing in the experiment schedule or in the database etc. In the recent years quantitative structure-activity relationship (QSAR) studies have shown several correlations of physiochemical properties and topological information indices. In the present paper, the use of an automatic structure generator is studied for detecting compounds with maximal activity by (carefully) extrapolating these correlations. So assumptions of optimal activity can be easier evidenced or discarded. The ideas are outlined on barbiturates and monoketones.