Molecular mechanics force fields are used to understand and predict a wide range of biological phenomena. However, current biomolecular force fields assume that parameters must be fit to the properties of small molecules and subsequently transferred to model large proteins. Here, we look to challenge this assumption and create a new class of QUantum mechanical BEspoke (QUBE) biomolecular force fields. QUBE is based around the use of atoms-in-molecule electron density partitioning to derive the non-bonded component of the force field. This thesis focusses on the derivation and validation of compatible bonded parameters that enable QUBE to be used in protein modelling.

Whilst parametrizing the bond and angle components of the new force fields, the inade- quacy of current parametrization schemes became apparent. This led to the development of a new bond and angle parametrization method that relies on only the quantum mechanical Hessian of a molecule. The new method resulted in the accurate recreation of the normal modes for a set of small molecules, heterocyclic molecules, dipeptides and a large osmium containing complex. The new method had an overall error in the normal mode frequency recreation of 6.3%, which is below that of the popular force field OPLS (7.4%).

Torsional parameters were also calculated for our protein force field and the conformational preferences of peptides and proteins were subsequently tested. Comparable accuracy to standard transferable force fields was achieved for simulations of short peptides, and this was demonstrated by the simulations’ J coupling errors, rotamer populations and backbone distributions. The J coupling errors remained at an acceptable level for protein simulations of ubiquitin and GB3, and two of the five proteins tested retained their experimental structure well during the MD simulations. In certain regions, particularly those with no clear secondary structure or a turn, three of the proteins exhibited some deviations from the experimental structure as the simulations progressed. However, given that this is the first generation of our QUBE force field, with future version envisaged, we view the results as promising. Additionally, improvements to the electrostatic potential of system-specific small molecule force fields were investigated. A new method was developed to add off centre point charges. The extra charges led to a reduction in the error of an atom’s electrostatic potential of 65.8%, as well as improvements to the free energy of hydration, for a benchmark set of molecules.

The methods and software developed in this thesis have the potential to improve the accuracy and accessibility of force field derivation, particularly for applications in biomolecular modelling.