Technical Highlight - January 2011
Short description: A detailed analysis of the HUP superfamily provides the groundwork for automated function prediction.
Proteins are composed of domains, and these protein domains can be grouped into superfamilies on the basis of their similarity. The size of these families varies considerably, and for some of the largest it seems that the protein domains within a family have developed diverse functions. Understanding how these differences arise might allow researchers to accurately predict functions.
This is the aim of a recent study by Christine Orengo and colleagues from PSI MCSG. They examine in detail one superfamily, the HUP domains. This family has a range of structures and functions: it includes various enzymes that catalyze reactions without a common mechanism, and it also includes non-enzymes.
The team classified by hand all the known structures according to the literature available on their function. Specific information, such as choice of substrate, was then used to further categorize domains within these subgroups. The results didn't fit within an already established classification framework and, instead, a new system had to be devised. This is a time-consuming approach and is not suitable for large-scale analyses.
The study shows that the main structural and functional differences within the HUP domain arose from 'embellishments', or additions, to the core structure, as previously suspected. In addition, from the new classification it was clear that certain residues were conserved within functional subgroups but not across the whole superfamily.
Using this information, the authors developed guidelines, which they then fed into an automatic protocol to classify domains with regard to their function. This protocol should be useful for classifying other large superfamilies, taking us a step nearer to accurately predicting function from structures or sequences.
B.H. Dessailly et al. Detailed analysis of function divergence in a large and diverse domain superfamily: toward a refined protocol of function classification.
Structure 18, 1522-1535 (2010). doi:10.1016/j.str.2010.08.017