Chapter 1 – A brief summary and overview of the history, development, and current approaches toward using quantitative structure-activity relationships in organic chemistry is presented. Examples of linear free-energy relationships and their use as inspiration for modern statistical modeling are shown. Using these statistical models, case studies are examined where mechanistic information is derived from the model, which is then used to explain reactivity and serve as a guide for new catalyst development. Finally, recent reports in the use of machine learning algorithms as a new approach in predictive model development are discussed.
Chapter 2 – Use of statistical modeling and mechanistic experiments for the understanding of the phosphine-driven, gold(I)-catalyzed [4+3]/[4+2] and [3+2]/[2+2] cycloisomerizations is described. Kinetic isotope effects and a linear correlation between the computed Au–Cl distance and [4+3]/[4+2] selectivity helps to demonstrate the phosphine’s electronic influence on product selectivity, while kinetic isotope effects, Hammett analysis, and a linear correlation between Sterimol L/B1 average and [3+2]/[2+2] selectivity suggest sterically-controlled product bifurcation. Insights from the [3+2/[2+2] model led to the development of a new ligand, AZPhos, which not only demonstrated higher selectivity in the [3+2]/[2+2] cycloisomerization, relative to the training set, but also showed utility in the arylation of 1,6-enynes, leading to higher product selectivity over the previously reported system.
2
Chapter 3 – The integration of statistical modeling for the development of higher selective methionine bioconjugation probes is described. An oxaziridine library is disclosed and is used to probe structural features controlling N- vs. O-transfer selectivity. Initial models led to the identification of outliers, prompting parameter collection and model development inspired by the Boltzmann distribution. A new, 4-parameter model was identified as well as a submodel that correlates the oxaziridine carbonyl stretching frequency (C=O) to the observed N- vs. O-transfer selectivity, leading to the identification of the t-butyl urea oxaziridine as the reagent giving the best N- vs. O-transfer selectivity. Alkynyl, t-butyl urea oxaziridines were synthesized and the translatability of the higher selectivity observed in our initial assay to biological systems was demonstrated.
Chapter 4 – Building on the strategy outlined in Chapter 3, this chapter describes the utilization of statistical modeling in conjunction with methionine bioconjugation and 1H NMR kinetics to understand the stability of the methionine sulfimide adduct and to identify a new oxaziridine leading to higher stability conjugates. It was found that the piperidine moiety led to highly stable sulfimides and the stability was shown to translate to peptides in vitro. Drawing inspiration from this scaffold, a new peptide stapling reagent was synthesized and was shown to effectively staple peptides at varying amino acid distances. The stability of the stapled adduct showed even greater stability compared to the unstapled counterpart in vitro, demonstrating how macrocyclization leads to increased stability. Finally, stapled and unstapled peptides with a fluorophore attached were synthesized and increased cell uptake for the stapled peptide was demonstrated in vivo.