The Atom Surface Site Interaction Point Approach to Non-Covalent Interactions

Advances in theoretical chemistry allowed us to discover the equations that describe the electronic structure of the atoms but also informed us that such accurate calculations are very expensive indeed. To look at larger supramolecular or biomolecular systems, or to investigate the effects of solvent, simplifications needed to be introduced that compromise the quality of the result, and the henceforth arising errors ought to be carefully considered. Furthermore, for many molecular design campaigns, the aim is not merely in learning about a single system, but the discovery of one with desirable properties, hence we need a reliable way of canvassing chemical space to find a molecule of interest. We have used DFT to develop the Atom Surface Site Interaction Point (AIP) approach to quantify all non-covalent interactions (NCIs) a molecule is capable of making with the environment, including solvent or any other binding partner in the liquid phase. The approach consists of an AIP description of molecules and a function that considers interactions between them. The description relies on linear fits between molecular electrostatic potential values and experimen tally determined hydrogen bonding parameters and has been extended to describe the molecules whole. The description has been tested with the free energy of transfer from n-hexadecane to water and showed good systematic prediction of phase transfer (RMSD of 5.2 kJ mol^-1). To consider interactions between two small molecules, we developed a custom pairing algorithm which identifies AIPs that are interacting at molecular interfaces and calculates the free energy of binding by summing over all pairwise interactions. Each such interaction has associated electrostatic and non-polar components and includes desolvation free energy. The method was optimised using experimentally measured free energies for aromatic interactions to test the predictive power of the approach in a simple well-defined system. The method is fast and produces interpretable binding free energies, as well as a graphical representation of the NCIs as Interaction Maps, that allow interpretation of the key contributions to the binding. The method has been applied to host-guest complexes in a variety of solvents. We also investigated protein-ligand scoring to assess the usability of the AIP approach in drug discovery. Using the CASF framework and dataset of diverse X-ray protein-ligand structures, we evaluated the AIP approach as a scoring function. It performs on par with the best available functions in the field and, furthermore, gives physical and accurate estimates of free energies of binding (RMSD of 10.3 kJ mol^-1). For a protein-ligand system of many ligands against a single EPHB4 protein, the AIP approach performed comparably to state-of-the-art free energy perturbation calculations. Therefore, the knowledge of non-covalent interactions obtained using the AIP approach can be used not only to judge promising ligand leads but also to design better binding ligands overall.

Name dc:type.qualificationname

Doctor of Philosophy (PhD)

Level dc:type.qualificationlevel

Doctoral

Grantor dc:publisher.institution

University of Cambridge

Year dc:date.issued

2024

Author dc:creator

• Zator, Katarzyna

Advisor dc:contributor.advisor

• Hunter, Christopher

• AIP
• DFT
• NCI
• Non-covalent interactions
• Protein-ligand
• SSIP
• Supramolecular

Licence

• https://apollo8-f-pro.lib.cam.ac.uk/bitstreams/f9d0f9cf-a237-41e0-b3f3-a8b79a3bffff/download
• https://creativecommons.org/licenses/by/4.0/

Language dc:language

eng

DOI dc:identifier.doi

https://doi.org/10.17863/CAM.112091

OAI identifier oai:identifier

oai:www.repository.cam.ac.uk:1810/373815