Development of a Novel Structure-Based Method

Virtanen, Salla Virtual screening: Development of a novel structure-based method Jyväskylä: University of Jyväskylä, 2013, 51 p. (Jyväskylä Studies in Biological and Environmental Science ISSN 1456-9701; 264) ISBN 978-951-39-5340-9 (nid.) ISBN 978-951-39-5341-6 (PDF) Yhteenveto: Virtuaaliseulonta; uuden rakennepohjaisen menetelmän kehitys Diss. Computational methods have become an integral part of today’s drug discovery and drug design efforts. The computational methods commonly used in virtual screening (VS) can be classified into two categories: ligand-based methods, which use information of known ligand molecules, and protein structure-based methods, which require information of the target protein structure. However, the currently used VS methods have some disadvantages: ligand-based methods tend to discover molecules that are very similar to the existing ligands, and protein structure-based methods often employ approximated scoring functions, which can lead to unreliable results. In this thesis a novel method for VS was developed, in which the shape of the ligand binding site is depicted as a ligand-like negative image. This negative image can then be used in place of a real active molecule in VS. When compared to traditional ligandand structure-based VS, the results with negative imagebased (NIB) screening were often better or comparable. Since the chemistry of the ligand binding site is also important for efficient binding, the chemical information was added to the negative images in the form of partial charges. The results show, that the addition of chemical information can improve the VS results, particularly in cases, where the binding site contains polar amino acids. Methods that employ implicit solvent models combined with molecular mechanics force fields, such as Molecular Mechanics Generalized Born Surface Area (MM-GBSA), have also become important tools in predicting binding free energies of ligands. We found that NIB screening can greatly benefit from postprocessing with MM-GBSA. The performance of MM-GBSA, along with Molecular Mechanics Poisson-Boltzmann Surface Area (MM-PBSA) and a similar method called Solvated Interaction Energy (SIE), was studied further, and our results show a highly case-specific performance both in VS and in predicting known binding affinities.