New molecular simulation methods for quantitative modelling of protein-ligand interactions
MetadataShow full item record
The main theme of this work is the design and development of new molecular simulation protocols, to achieve more accurate and reliable estimates of free energy changes for processes relevant to the structure-based drug design. The works starts with an insight into the reproducibility problem for alchemical free energy calculations. Even if simulations are run with similar input files, the use of different simulation engines could give different free energy results. As part of a collaborative effort, the implementation details of AMBER, GROMACS, SOMD and CHARMM simulation codes were studied and free energy protocols for each software were validated to converge towards a reproducibility limit of about 0.20 kcal.mol-1 for hydration free energies of small organic molecules. Following, new simulation methods for the estimation of lipophilicity coefficients (log P and log D) for drug like molecules were developed and validated. log P values were computed for a dataset of 5 molecules with increasing fluorination level. Predictions were in line with the experimental measures and the simulations also allowed new insights into the water-solute interactions that drive the partitioning process. Then, as part of the SAMPL5 challenge, log D values for 53 drug-like molecules were computed. In this context two different simulation models were derived in order to take into account the presence of protonated species. The results were encouraging but also highlighted limits in alchemical free energy modelling. As an additional task of the SAMPL5 contest, three different protocols were validated for predicting absolute binding affinities for 22 host-guest systems. The first model yielded a free energy of binding based on free energy changes in solvated and complex phase; the second added the long range dispersion correction to the previous model; the third one used a standard state correction term. All three protocols were among the top-ranked submission in SAMPL5, with a correlation coefficient R2 of about 0.7 against experimental data. Finally, the origins and magnitude of the finite size artefacts in alchemical free energy calculations were investigated. Finite size artefacts are especially predominant in calculations that involve changes in the net-charge of a solute. A new correction scheme was devised for the Barker Watts Reaction Field approach and compared with the literature. Hydration free energy calculations on simple ionic species were carried out to validate the consistency of the scheme and the approach was further extended to host-guest binding affinities predictions.