The development of hybrid quantum classical computational methods for carbohydrate and hypervalent phosphoric systems

Doctoral Thesis


Permanent link to this Item
Journal Title
Link to Journal
Journal ISSN
Volume Title

University of Cape Town

Ab initio, density functional theory, and semi-empirical methods serve as major computational tools for quantum mechanical calculations of medium to large molecular systems. Semi-empirical methods are most effectively used in a hybrid quantum mechanics/molecular mechanics (QM/MM) dynamics framework. However, semi-empirical methods have been designed to provide accurate results for organic molecules, but often fail to treat hypervalent species accurately due to their use of an sp basis. Recently, significant breakthroughs have been made with the incorporation of d-orbitals into the semi-empirical framework, thereby allowing for accurate modeling of both hypervalent and transition metal systems. Here I consider two methods that adopt this new methodology, namely AM1/d-PhoT and AM1*. Our major focus is the simulation of chemical biological and more specifically chemical glycobiological problems of biochemical interest. When I tested the ability of both AM1/d-PhoT and AM1* to reproduce key metrics in chemical glycobiology (i.e., sugar ring pucker, phosphate participation in transferase reactions) these methods, in combination with the published parameters, performed very poorly. Using the AM1/d-PhoT and AM1* Hamiltonians I set out to re-parameterize these methods aiming to produce holistic biochemical QM/MM toolsets able to simulate fundamental problems of binding and enzyme reactivity in chemical glycobiology. We called these methods AM1/d-CB1 and AM1*-CB1. In the development of these parameter sets I focused specifically on proton transfer, carbohydrate ring puckering, bond polarization, amino acid interactions, and phosphate interactions (facets important to chemical glycobiology). Both AM1/d-CB1 and AM1*-CB1 make use of a variable property optimization parameter approach for the glycan molecular class and its chemical environment. The accuracy of these methods is evaluated for carbohydrates, amino acids and phosphates present in catalytic domains of glycoenzymes, and the are shown to be more accurate for key performance indices (puckering, etc.) and on average across all simulation derived properties (QM/MM polarization, protein performance, etc.) than all other NDDO semiempirical methods currently being used. A major objective of the newly developed AM1/d-CB1 and AM1*-CB1 is to provide a platform to accurately model reactions central to chemical glycobiology using hybrid QM/MM molecular dynamics (MD) simulations. AM1/d-CB1 is applied to a well-known reaction involving purine nucleoside phosphorylase (PNP) and results lead me to conclude that the method shows promise for modelling glycobiological QM/MM systems.

Includes bibliographical references.