During the XX century the dominant paradigm in creation of new drugs was based on suggestion about selectivity of action on a certain molecular target, which should lead to the normalization of pathological process. At the same time, the majority of known drugs interact with several or even many targets in the organism; however such action is mostly associated with unwanted adverse effects and toxicity. After the deciphering of human genome and first results achieved in postgenomic studies it became obvious that many diseases have a complex etiology, while drug action on a certain target often leads to activation/blockade of other elements in the appropriate regulatory network. As a consequence of negative feedbacks, expected pharmacotherapeutic action may be significantly decreased or even completely suppressed. The multitargeted drugs concept appears, according which such remedies due to the additive, synergistic or antagonistic action might have some advantages comparing to the monotargeted medicines. Discovery of new multitargeted drugs requires the solution of two tasks: (1) identification of targets which could be blocked or activated to achieve the desirable pharmacotherapeutic effect, and (2) finding of ligands that interact with the identified targets in a desirable mode. The first aim could be achieved by simulation of behavior in regulatory network blocking or activating certain nodes (targets) and their combinations. The second aim could be achieved either by prediction of biological activity spectra for molecules from different databases and selection of hits with the requested biological profiles, or by direct computer-aided design of multitargeted ligands. Possibilities and limitations of current bioinformatics and computer-aided drug design methods in discovery of particular molecular targets and their multiple ligands will be discussed.
Med. Chem. Res., 2010, 19 (S1), s30.

Rational selection of potential antineoplastic agents with multitargeted action is considered on the basis of computer prediction of biological activity spectra using PASS program.
Chemistry Central Journal, 2008, 2 (Suppl.1), 18.

The prediction of biological activity spectra for substances as an approach for searching compounds with complex mechanisms of action was studied. New compounds with dual mechanisms of antihypertensive action were found by this approach. Biological activity spectra for substances were predicted on the basis of their structural formulas by the computer program PASS. Thirty molecular mechanisms of action of compounds from the MDDR 99.2 database, which cause the antihypertensive effect and can be predicted by PASS, have been identified. The analysis of predictions for compounds with 15 dual antihypertensive mechanisms of action from the MDDR 99.2 database has confirmed high accuracy of prediction. This approach was applied to databases of commercially available compounds (AsInEx and ChemBridge) and allowed us to select four substances that are potential inhibitors of angiotensin converting enzyme (ACE) and of neutral endopeptidase (NEP). At a later time, all these compounds were found to be the inhibitors of both ACE and NEP. The most potent compounds had IC50 of 10-7 - 10-9 M for ACE and 10-5 M for NEP. New combinations of dual mechanisms of action never before found for antihypertensive compounds were predicted.
Journal of Medicinal Chemistry, 2003, 46 (15), 3326-3332.