Assessment of the Efficiency of Selecting Promising Compounds During Virtual Screening Based on Various Estimations of Drug-Likeness

Progress in synthetic and medicinal chemistry significantly expands the chemical space where new pharmacological substances are discovered. Virtual screening is a preliminary selection of chemical compounds potentially possessing the desired properties that is carried out in silico based on an assessment of the structural similarity, pharmacophore analysis, (Q)SAR, and molecular modeling. The use “filters” is proposed to assess the compounds for their “Drug-Likeness” properties, i.e., the resemblance of the analyzed compounds to known drugs, to increase the probability of identifying compounds with the desired properties. We developed the World-Wide Approved Drugs (WWAD) database of pharmaceutical substances approved for medical use in 52 countries, which allows us to evaluate the pros and cons of these filters. It is shown that applying the rules of Lipinski, Ghose, Weber, REOS, and QED to a set containing about a million chemical compounds from the PubChem database and more than 4000 chemical compounds from the WWAD database does not distinguish “drugs” from “non-drugs” and leads to the exclusion of a significant part of registered drugs from further studies. The application of the Prediction of Activity Spectra for Substances (PASS) software package appeared to be the most effective approach to assess the Drug-Likeness.