Fragment-based drug design integrates different methods to create novel ligands using fragment libraries focused on the particular biological activities. Experimental approaches to fragment libraries preparation have some drawbacks caused by necessity of target crystallization (X-ray and NMR) and careful immobilization (SPR). Molecular modeling (docking) requires accurate data on protein-ligand interactions, which are difficult to obtain for some proteins. The main drawbacks of QSAR application are associated with the need to collect large homogeneous datasets of chemical structures with experimentally determined self-consistent quantitative values (potency). We propose the ligand-based approach to the selection of fragments with positive contribution to the biological activity, developed on the basis of the PASS algorithm. The robustness of PASS algorithm for heterogeneous datasets has been shown earlier. PASS estimates qualitative (“yes”/”no”) prediction of biological activity spectra for over 4000 biological activities and, therefore, provides the basis for preparation of fragment library corresponding to multiple criteria. Algorithm of fragments selection has been validated using the fractions of intermolecular interactions calculated for known inhibitors of nine enzymes extracted from PDB. The statistical significance of differences between fractions of intermolecular interactions corresponds, for several enzymes, to the estimated positive and negative contribution of fragments into the enzyme inhibition.
SAR & QSAR Environ. Res., 2012, ., 23 (3-4), 279-296