Development of reactive force fields and empirical scoring functions for drug screening

Ion-peptide interactions play crucial roles in many systems, for example, in neurodegenerative diseases. An accurate computational model can assist in understanding these biologically relevant systems, as well as designing bio-mimetic sensors that can bind particular metal ions with high sensitivity and selectivity. This dissertation describes the development of a reactive force field, ReaxFF, that is a first step toward achieving this goal. The ReaxFF force field has been optimized against a quantum mechanics based training set that accurately describes the physical chemical properties of Cu2+, Cl- and CuCl2 in the gas phase and in aqueous environments. These potentials can successfully reproduce the coordination geometries and energies of several systems previously studied by experimental and theoretical means. These potentials even reproduce purely quantum effects like Jahn-Teller distortion. A ReaxFF potential was also developed for glycine in the gas phase and in water. This potential was successfully applied to simulate the tautomerization of glycine as it enters the aqueous phase from the gas phase. These potentials can be extended to include other amino acids and small organic molecules. The potentials developed in this work provide a computationally inexpensive tool for large scale dynamic simulations of reactive systems with good accuracy. They can assist in understanding the complex properties of these systems as well as design novel tools with diverse applications.;Three novel scoring functions for drug screening have been developed with computational chemistry methods. The empirical scoring functions were trained and validated against a large and diverse set of protein-ligand complexes. The performances of some of the scoring functions were comparable to the performance of accurate force field based linear interaction energy method in terms of predicting binding free energy, distinguishing native against decoy poses and enriching actives from decoys in virtual screening process. These scoring functions have potential applications in discovering and designing novel and more effective drugs. In general, the methods developed in this work facilitate understanding of a diverse set of chemical and biological systems at the molecular scale.