Computational predictions and reactivity analyses of organic reactions

Abstract

This thesis focuses on computational reactivity analyses and predictions for organic systems. The research began with studies on specific reactions using quantum mechanics and molecular mechanics simulations. Beyond looking at individual reactions, projects in method development were also initiated to set up a more efficient and automated procedure for tackling the challenges of exploring the conformational space, investigating the effect of reaction dynamics and suggesting possible reaction pathways. Chapter 2 focuses on the selective pyridination of 5-methylcytosine. A new reaction mechanism that aligns with the experimental result was proposed. The variations in percentage yield upon changes in the substituent in the pyridine substrate can be explained by considering the thermodynamics and kinetics of the proton transfer step from cationic 5mdC•+ to neutral pyridine substrates. Substrates with an electron-withdrawing group are disfavoured thermodynamically, while substrates with an electron-donating group face a high kinetic barrier in a key proton transfer step. Chapter 3 gives insights into the origin of enantioselectivity in 1,4-dicarbonyl synthesis reactions with diarylprolinol silyl ether catalysts via a radical pathway. A robust procedure has been developed for computational investigations of large and flexible chemical systems based on the conformation labelling system, ONIOM calculations and Python scripting. The change in enantiomeric excess due to variations in the catalyst can be explained based on conformational changes and structural deformations. In the enantioselectivity-determining radical addition step, the iminium in the lowest energy SR transition state (TS) takes up the conformation of the lowest energy ground state iminium (EE). The conjugated iminium in the SS TS adopts an EZ conformation to avoid potential structural deformations due to radical attacks from the more sterically hindered position. For systems with simpler catalysts, the iminium is EE for both, as the steric hindrance imposed by the substituent is not sufficient to cause this large structural deformation, and so the reaction shows poor enantioselectivity. Chapter 4 is on the development of the CONFPASS (Conformer Prioritisations and Analysis for DFT re-optimisations) Python package. CONFPASS extracts dihedral angle descriptors from conformational searching outputs, performs clustering, and returns a priority list for density functional theory (DFT) re-optimisations. Evaluations were conducted with DFT data of the conformers for 150 structurally diverse molecules, most of which are flexible. CONFPASS gives a confidence estimate that the global minimum structure has been found, and based on our dataset, we can have 90% confidence after optimising half of the FF structures. Re-optimising conformers in order of the FF energy often generates duplicate results; using CONFPASS, the duplication rate is reduced by a factor of two for the first 30% of the re-optimisations, which includes the global minimum structure about 80% of the time. Chapter 5 details the study of a group of oxazaborolidinium ion-catalysed reactions between aldehyde and diazo compounds. Their selectivity cannot be explained using transition state theory. VRAI-selectivity, developed to predict the outcome of dynamically controlled reactions, can account for both the chemo- and the stereo-selectivity in these reactions, which are controlled by reaction dynamics. Subtle modifications to the substrate or catalyst substituents alter the potential energy surface, leading to changes in predominant reaction pathways and altering the barriers to the major product when reaction dynamics are considered. In addition, this study suggests an explanation for the mysterious inversion of enantioselectivity resulting from the inclusion of an ortho iPrO group in the catalyst. Chapter 6 presents a method for predicting reaction sites based only on a simple, two-bond model. Machine learning classification models were trained and evaluated using atom-level labels and descriptors, including bond strength and connectivity. Despite limitations in covering only local chemical environments, the models achieved over 80% accuracy even with challenging datasets that cover a diverse chemical space. Whilst this simplistic model is necessarily incomplete, it describes a large amount of interesting chemistry and provides guidance on potential reaction pathways.