Computing Aqueous Solubility and Understanding Hydrophobicity
James McDonagh
The project will develop new methods of calculating the aqueous solubility of organic molecules. Its philosophy is to compute all the relevant solute-solute, solute-solvent, and solvent-solvent enthalpy and entropy terms. This must be done both for the initial state, pure water and the solute in its lowest free energy crystalline state, and also for the final state, the solution. We have previously developed models that combine a theoretically rigorous treatment of the crystalline state with a description of the aqueous solution that incorporates empirical structure-property relationships. We intend to expand on this work, widening its applicability, and especially to develop its biochemical relevance. In particular, we will develop a model that is capable of systematic improvement by partitioning the free energy of solvation into a number of distinct and physically meaningful enthalpic and entropic components. This will include more accurate computation of the solute-solvent interaction.
To date, we have produced a purely theoretical method, which has not been parametrised on any solubility data. This method makes good solubility predictions for a 25 molecule dataset, which we used as a proof of concept. Following from this we produced several QSAR models of solubility, utilising machine learning methods. Currently, our interest has focused on enhancing our theoretical methodology, utilising periodic DFT to model crystal structures. In addition to these main projects, I have briefly investigated melting point prediction and have an interest in crystal structure prediction.
Its primary practical applications involve in silico virtual screening of molecules for drug discovery. The work will also be of value, however, in improving our understanding of the molecular basis of the hydrophobic effect. Such knowledge can be used in diverse areas, for instance modelling protein folding, predicting the blood-brain partitioning of pharmaceuticals, and computing protein-ligand binding affinities.
This work was presented at the ScotChem 2013 and can be found at the following link James ScotChem 2013. A video of the presentation can be found here .
Some of the scripts and dataset used in our work can be found here.
A current CV is available for download.
Contact information
Email: jm222@st-andrews.ac.uk
Phone: 01334 464748 ext: 14141
James McDonagh's publications and presentations
