Function prediction using patches, pockets and other surface properties

Abstract

With few exceptions protein functions depend sensitively upon their interactions with other biomolecules. Thus, the surface of a protein is of particular interest for function annotation: definition of the protein surface in experimental or modelled protein structure enables the application of a wide range of structural bioinformatic tools for function prediction. The development of such tools has been significantly accelerated in recent years as a response to the flux of information from Structural Genomics programs which, at least in part, have deliberately targeted mysterious protein families of unknown function about which conventional homology-based protein function annotation can say little or nothing (Bateman et al. in Acta Crystallographica Section F: Structural Biology and Crystallization Communications 66:1148-1152, 2010). As this chapter will illustrate, the underlying principles behind the resulting toolset vary impressively but, ultimately, most are based upon discovering putative interaction sites through detecting ways in which they differ somehow from protein surface in general. These differences may be physicochemical, electrostatic or steric in nature, or be of evolutionary origin. Predictions can be strengthened by observing concordant results from orthogonal methods. Indeed, many programs now improve performance by combining multiple factors in their calculations. Some methods find functional sites in general, others provide direct evidence supporting specific biochemical functions. This chapter will not attempt a comprehensive historical overview of the area, rather aiming to guide the user to the current state of the art while acknowledging key methodology papers. Methods that are readily available will be favoured, particularly those implemented at servers and those for which plug-ins for popular molecular visualisation tools exist.