Current approaches to drug discovery - new directions and unavoidable difficulties. Recent views on a slow-down in new active substance discovery are discussed. Drug discovery paradigm based on observation of an advantageous action of a specific substance as starting point for further development is compared with the current approach based on mining of huge libraries of substances produced by combinatorial chemistry methods and tested using high throughput screening techniques. The use of the molecular modeling in the search for receptor ligands and the role of adequate molecular skeletons are reviewed.