Source: Dissertation Abstracts International, Volume: 79-11(E), Section: B.

Adviser: William L. Jorgensen.

Computational structure based drug discovery has become a promising approach for the development of new small molecule drugs. Improved force fields are developed for proteins and RNA, yielding higher accuracy results for simulations of such systems. Computational approaches were then applied to discover high-potency inhibitors of macrophage migration inhibitory factor. These compounds were then subjected to extensive crystallography to validate computational predictions and guide further optimization.

The development and validation of new peptide dihedral parameters are reported for the OPLS-AA force field. High accuracy quantum chemical methods were used to scan ϕ,psi chi1, and chi2 potential energy surfaces for blocked dipeptides. New Fourier coefficients for the dihedral angle terms of the OPLS-AA force field were fit to these surfaces. Extensive experimental solution-phase and quantum chemical gas-phase benchmarks were used to assess the quality of the new parameters, named OPLS-AA/M, demonstrating significant improvement over previous OPLS-AA force fields.

Next, in this work, several popular force fields are evaluated for reproducing experimental properties of the flavodoxin/flavin mononucleotide system. Free energy perturbation calculations were also executed between different protein mutants for comparison with experimental data for relative free energies of binding. OPLS-AAIM paired with CM5 charges for the ligand performed particularly well, both for the 3J couplings and FEP results, with a mean unsigned error for relative free energies of binding of 0.36 kcal/mol.

DFT calculations have been used to develop improved descriptions of the torsional energetics for nucleosides and nucleotides in the OPLS-AA force field. Scans of nucleotide dihedral angles (gamma, chi, and beta and methyl phosphates provided the bases for the new torsional parameters. In addition, the angle-bending parameters of phosphodiesters and ribose were updated, and adjustments were made to existing carbohydrate torsions to better capture the sugar puckering landscape of ribose. MD simulations of nucleosides with the new parameters demonstrate a significant improvement in the ribose sugar puckering and chi angle distributions.

Computational and crystallographic studies involved in the optimization of biaryl triazole inhibitors of macrophage migration inhibitory factor are reported. Co-crystal structures of several biaryl triazole inhibitors are reported revealing a consistent binding mode. Relative free energy perturbations were perfoiuied to explore substitutions at several points in the molecule as well as small alterations to the scaffold. Potential sites for improvement were identified. A small molecule crystal structure of one of the inhibitors was solved, and the small molecule packing was examined to rationalize the impact of substitutions upon the solubility of the compounds.

While the biaryl triazole series was able to be optimized into a highly potent inhibitor of MIF, there is still a strong desire pursue other inhibitors of MIF. In the course of initial screening of compounds, pyrazoles were identified as a potential inhibitor, with an initial potency of 100 micromolar. An initial crystal structure was obtained, allowing for further optimization of the compound. Successive rounds of synthesis and crystallography were able to obtain low micromolar binders. Free energy perturbation calculations were used to examine the impact of fluorination of the pyrazole, suggesting a significant improvement in binding affinity and a specific bound pose, both of which were validated experimentally.

There is a strong desire to provide quantitative examinations of the accuracy of computational drug discovery methods. However, demonstration that a computational method can accurately recapitulate the experimental data is meaningless if the experimental data is not accurate. In the course of our work on MIF, several literature compounds were purchased and assayed. Binding affinities in the literature for several compounds were inconsistent. Crystal structures were solved for several of these compounds bound to MIF to interrogate a potential structural basis for the discrepancy.

Orbital theory provides a powerful tool for rationalizing and understanding many phenomena in chemistry. In this work, we describe a general method for producing 3D printing files of orbital models that can be employed with most popular software packages for performing electronic structure calculations and molecular visualization. Numerous examples of various systems of interest in physical organic chemistry are provided in the .stl format for 3D printing, as well as a fully illustrated tutorial for the process.

Finally, significant improvements were made to the OPLS-AA force field for RNA. Quantum chemical scans of the alpha/gamma potential energy surface were performed and new parameters were fit for the corresponding torsion potentials. The new force field was validated with diverse molecular dynamics simulations. Results for dinucleotides and tetranucleotides reveal accurate reproduction of 3J couplings from NMR. Simulations of common noncanonical motifs reveal improvement over the previous OPLS-AA force field.