Case Studies In Assigning Function From Structure In Structural Genomics

James Watson
European Bioinformatics Institute

A significant proportion of the protein structures solved by the various Structural Genmics groups are of hypothetical proteins of unknown function. As part of the Midwest Centre for Structural Genomics (MCSG) we have developed a fully-automated structural analysis and functional annotation server. The server, called Profunc, uses a wide range of different methods including sequence analysis, residue conservation scoring, fold recognition, surface cleft analysis, active site template searching and genomic location analysis. Although the analyses are run automatically the results inevitably require significant human interpretation and assessment. To this end, a small dataset comprising 63 non-homologous MCSG structures was submitted to Profunc and the functional assignments examined. Our analysis suggests that the sequence based information is still the most successful in assigning function but that in a number of cases the use of fold homology and template-based methods can significantly improve the assignment of function. The analysis also suggests that in almost 25% of Structural Genomics targets the current methods will be unable to assign any function. A larger dataset of over 350 Structural Genomics deposits is currently being examined to assess this figure and to determine how successful each method is.

Presentation (PowerPoint File)

Back to Long Programs