The working group deals with the use of computer technology in the development of a wide range of potential drugs. Among the routinely activities we perform primarily molecular mechanics and quantum chemical calculations. The most important molecular mechanical tasks are represented by flexible molecular docking, molecular dynamics and structure-based virtual screening. Furthermore, we also carry out conformational analysis of chemical models at the level of molecular mechanics and semi-empirical computational chemistry methods. The results obtained on a lower level of theory are used as initial data for demanding quantum chemical calculations. From the set of available quantum chemical calculation methods we apply especially those which are based on the electron density functional theory (DFT), especially hybrid variants combining Beck functional electron exchange energy with the exact energy of the Hartree-Fock Theory (e.g. B3LYP). The resulting quantum-chemical models are used especially to derive quantum-chemical descriptors and, subsequently, to study quantitative relationships between the structure and activity (i.e. QSAR). For calculations of quantum chemical descriptors, classical QSAR, 3D QSAR analyses, etc., we develop our own algorithms and computer programs.
Data gathered in the phase of chemical calculations are analyzed by various data mining tools, tailored to the needs of bioinformatics. For treating extensive information on potential drugs we employ a custom MySQL database, and this data is processed for example by methods of ligand-based virtual screening. Filtering the virtual ligand database is performed for instance according to Lipinski's rules in combination with the search for structural similarity relative to the selected template substance (i.e. a lead structure). Other methods of data mining are represented by statistical tools such as multiple linear regression (MLR), partial least squares regression (PLS), principal component analysis (PCA), cluster analysis (CA) and artificial neural networks (ANN). Acquired statistical models are appropriately validated by cross-validation and prediction on an external set, and eventually used in the design of new potentially bioactive structures.
Examples of used software:
Gaussian, Matlab, Statistica, AutoDock Vina, PyMOL, Schrödinger, HyperChem, Spartan, and others.
Cooperating computer centers:
FIM UHK, MetaCentrum, IT4Innovations.