%0 Journal Article
%T Partially-supervised protein subclass discovery with simultaneous annotation of functional residues
%A Benjamin Georgi
%A J£¿rg Schultz
%A Alexander Schliep
%J BMC Structural Biology
%D 2009
%I BioMed Central
%R 10.1186/1472-6807-9-68
%X We have developed an extension of the context-specific independence mixture model clustering framework which allows for the integration of experimental data. As these are usually known only for a few proteins, our algorithm implements a partially-supervised learning approach. We discover domain subfamilies and predict functional residues for four protein domain families: phosphatases, pyridoxal dependent decarboxylases, WW and SH3 domains to demonstrate the usefulness of our approach.The partially-supervised clustering revealed biologically meaningful subfamilies even for highly heterogeneous domains and the predicted functional residues provide insights into the basis of the different substrate specificities.Protein families frequently can be divided into subfamilies of similar but distinct function. The study of these subfamilies and the residues which control the functional specificity is an important step in the analysis of these families.Many previous studies have focused on the question of how to find the functional residues for a given protein family when proteins already have been assigned to subfamilies. These methods include approaches based on information-theoretical measures such as relative entropy [1,2] or mutual information [3], template-based similarity scores to known functional residues [4], approaches which contrast position-specific conservation in orthologues and paralogues [5] or superfamilies [6] and comparisons to known reference 3D structures to find discriminatory surface residues [7]. The opposite of this so called supervised problem, is the unsupervised setup, where subfamily assignments are unknown and have to be inferred from the data. In the unsupervised case, clustering approaches for protein data can be applied to obtain subfamilies from set of protein sequences. For protein subfamily clustering most methods rely on the construction of a phylogenetic tree. These methods can be further subdivided into pure clustering methods [8-10] an
%U http://www.biomedcentral.com/1472-6807/9/68