Protein solvation preferences: applications to drug discovery

Computer-aided drug design is a key player in current drug discovery projects. Structure-based computational approaches use the target structural information to suggest potentially active and safe drugs. However, the process is far from trivial and novel methodologies are continuously sought to address two main factors usually simplified and overlooked: Target flexibility and the effect and structure of water molecules at the binding site.\n\n\t\t\t\t As demonstrated by different NMR and crystallography experiments, small organic solvents (e.g. ethanol, isopropanol, acetonitrile) are able to identify binding sites and provide clues for rational drug design. MDmix is a simulation-based method that exploits this natural behavior in silico. By using small organic molecules and water mixtures, each one with a distinct chemical nature, key interaction spots are identified on the protein surface allowing the identification and characterization of binding sites for hit discovery and lead optimization. \n\t\t\t\t The work presented in this thesis is divided in two main publications: In the first one, the effect of target flexibility was investigated to establish some guidelines on how to treat this important factor during the simulations. We found that flexibility is essential for correctly identifying induced binding sites but might lead to uninterpretable results when large conformational changes occur. Soft restraints applied during the simulation are suggested as a way to obtain reproducible results and still characterize high affinity interaction sites (hot spots) with mild errors on the energy estimates. \n\n\t\t\t\t In the second publication, the use of solvent mixtures for the identification of experimentally known pharmacophores was evaluated in two test systems for which many inhibitors are known (e.g. heat shock protein 90 and HIV protease 1). The explicit treatment of water molecules provides interaction maps which identify the most favorable interactions in the binding site with unprecedented accuracy when compared to classical molecular interaction potentials. Moreover, we demonstrate how the interaction maps obtained for the water molecules accompanying the small organic solvents are useful to identify non-displaceable waters. Both the solvent interaction maps and the water interaction maps are extremely useful information for the identification of novel active molecules and for the optimization of potency for already active ones. \n\n\t\t\t\t Finally, a software package is presented that aims at facilitating the use of the methodology and at helping in adopting it to everyday drug design projects. A final chapter treats ongoing and future research where method improvements and practical uses in real examples are discussed.\n\n\t\t\t\t MDmix being a simulation-based method, the target flexibility and the explicit treatment of the solvent provide significant advantages over traditional approaches for binding site finding and characterization. This novel approach, which is applicable to previously unmet targets and binding sites, offers a new alternative in the challenging process of drug design.