Paul Bates

Computational approaches to the construction and analysis of protein networks

Computational approaches to protein network analysis are important if we are to gain a deeper understanding of cellular processes, such as signal transduction, DNA repair and the immune response.

Protein network analysis is primarily dependent on reliable assignment of protein-protein interactions. Several different experimental methods are currently in use to detect such interactions. Two-hybrid screens are, to date, the approach that has yielded the most volume of data. However its level of accuracy is not particularly high, unless supported by additional evidence - further experimental and/or computational.

Knowledge of protein structure can assist in confident protein-protein interaction assignment. For example, protein pairs can be docked together using computer based algorithms¹. However, one of the major problems we face is that three-dimensional coordinates are not yet available for all proteins, hence many must be modelled. This is usually achieved by comparative modelling, a technique that is expected, along with structural genomic projects, to eventually provide complete coverage of three-dimensional protein folding space.

The graduate student will be able to take advantage of our continual development of computer-based algorithms aimed at improving the accuracy of comparative modelling².

The student will analyse the computer generated protein models and assign properties to them that indicate which potential binding partners they are likely to interact with. This will require the use of such tools as sequence conservation mapping to protein surfaces. These potential binding sites will then be used as restraints in our protein docking algorithms¹.

Once confident protein interaction networks are established, and the focus here will be networks particularly associated with cancer³, the fluctuations and interdependencies of the networks will be studied by subsequent mapping of expression data, especially data derived from microarray experiments. The properties of some of these networks, and associated molecular pathways, will be further investigated by running our cell-based simulation software⁴.

References

1. Krol M, Tournier AL and Bates PA. Flexible relaxation of rigid-body docking solutions. Proteins: Structure, Function and Bioinformatics
2007; 68: 159-169.
2. Offman MN, Fitzjohn PW and Bates PA. Developing a move-set for protein model refinement. Bioinformatics
2006; 22: 1838-1845.
3. Jonsson PF and Bates PA. Global topological features of cancer proteins in the human interactome. Bioinformatics
2006; 22: 2291-2297.
4. Tournier AL, Fitzjohn PW and Bates PA. Probability-based model of protein-protein interactions on biological timescales. Algorithms Mol Biol
2006; 1: 25.