Most proteins are organized in domains which can be seen as independent modular units in terms of molecular function (MF). Nevertheless, current functional annotations are done on a "whole-chain" basis without associating specific functions to the individual domains. We present here an automatic method for discerning which particular structural domain within a protein is responsible for a given MF originally attributed to the whole protein. By annotating the SCOP structural domains with gene ontology terms using this method, we obtained the first large-scale functional annotation at the domain level. We performed a large-scale comparison of these annotations with the ones implicit in the functional annotations of Interpro signatures, showing that the performance of this method is globally better. We also discuss in detail some particular examples. Generated automatically and available online, this resource could be the basis for future manually curated annotations.
2009 Wiley-Liss, Inc.