Matthew Seddon

Location: Information School, Regent Court - Michael Lynch Laboratory (RC-323)

Seddon 100x125


Project title: Development of novel techniques for assessing bioisosteric similarity of chemical fragments
Supervisors: Professor Val Gillet (Information School), David Cosgrove (AstraZeneca) and Martin Packer (AstraZeneca).

Project description: Bioisosterism relates to similarity in biological activity between two compounds. It is often assumed that when two molecules are structurally similar that they will have a similar biological function. However, there are known cases of two molecules with a similar activity that are not structurally similar. Current state of the art similarity methods are focused on structural similarity. This project aims to find a similarity method that would be structure invariant, so that it would discover molecules that are of similar biological activity that are not necessarily structural ananlogues. This will be achieved through the use of advanced 3D shape and field based methods or through quantum mechanical representations of the molecules.
Currently these methods are inefficient on large molecules. Molecules are flexible, often having more than one low energy conformation in solution, and so quantifying these properties on large molecules is computationally expensive and often results in ineffective representations. Fragment Based Drug Discovery, the approach of building drug molecules by identifying functional small fragments, avoids many of these problems. Therefore, this project aims to apply these methods to fragments. Preliminary analysis has focussed on analysing the fragments of crystallographic data and the nonlinear similarity properties of molecules. Future work will investigate new computational geometry methods using spectral and diffusion geometry along with methods from computational geometry. These methods take the spectrum of the Laplace Beltrami operator over the surface of a 3D object to compute intrinsic shape descriptors.

Research interests: For my project I am interested in computational geometry methods for molecular shape. In particular, spectral and diffusion geometry for isometric shape description. This incorporates areas of computational geometry and algebraic topology. I have a general interest in chemoinformatics, computational drug design, statistics, machine learning, and probabilistic modelling.

Achievements and awards:

  • Chair - iFutures Committee 2015
  • UKQSAR Spring 2016 - best poster prize for "Application of spectral and diffusion geometry to 3D molecular surface description

Journal articles