Building meaningful atomic models of polysaccharides automatically

Building meaningful atomic models of polysaccharides automatically


Project Description

We have recently assessed the methodological support for building atomic models of carbohydrates using data from macromolecular crystallography (MX) and electron cryo-microscopy (cryoEM), and found it to be very incomplete. As part of our analysis, we have found thousands of unlikely or indeed wrong carbohydrate structures in public macromolecular structure databases. One of the widely-accepted reasons for this is the lack of automated tools for creating and fitting atomic models of carbohydrates to the experimental data. This is a problem now for MX, and it is likely to become a major problem for cryoEM as the technique becomes ever more popular.

We have an evolving set of high-performance tools - e.g. Privateer - which can be used as building blocks for creating a much bigger and capable computer program. While our software is built on efficient but harder to maintain languages (C++ with use of parallelism), we will use the high-level, easy to learn Python programming language to combine pre-existing modular code in logical ways. The ultimate goal is to build a computer program that combines all our knowledge of carbohydrate chemistry and glycobiology to create meaningful atomic models of branched polysaccharides. The tool will be evaluated against all published protein-carbohydrate structures - the results are expected to enhance or even increase our glycobiological insight.

This project is framed within the EPSRC Research Theme 'Physical sciences', covering several research areas such as Artificial intelligence, Biological informatics,Chemical biology and biological chemistry, Computational and theoretical chemistry, Databases, Digital signal processing, and Graphics and visualisation.