Functional Annotation of Proteins through Substructure Matching

The number of known protein structures is rapidly increasing. The function of most proteins is still poorly understood or even completely unknown. At the same time, there are also many large protein families for which many structural variants are available. We present a substructure-based approach called LabelHash that can be used to annotate proteins with unknown function. We also describe a new method that uses LabelHash as a tool to help understand the structural variations within classes of proteins with known function. This structural variation within a family of related proteins can be exploited to design drugs with very high specificity. Determining the function of unannotated proteins would have a significant impact on understanding diseases and designing new therapeutics. However, experimental protein function determination is expensive and very time-consuming. Computational methods can facilitate function determination by identifying proteins that have high structural and chemical similarity. Our focus is on methods that determine binding site similarity. Although several such methods exist, it still remains a challenging problem to quickly find all functionally-related matches for structural motifs in the entire Protein Data Bank (PDB) with high specificity. In this context, a structural motif is a set of 3D points annotated with physicochemical information that characterize a molecular function. We have developed a method called LabelHash that creates a hash table of n-tuples of residues for all structures in the PDB [2]. The method is inspired by geometric hashing, a technique that originated in computer vision but which has also been applied to matching structural motifs [6, 5]. The key advantage of LabelHash over geometric hashing is that it uses much less space and scales more easily to very large data sets such as the entire PDB. Using the LabelHash hash tables, we can quickly look up partial matches to a motif and expand those matches to complete matches. We show that by applying only very mild geometric constraints we can find statistically significant matches with extremely high sensitivity and specificity for very general structural motifs (see Figure 1). The LabelHash method is also extremely fast; it can match motifs ranging in size from 3 to 11 residues in a matter of seconds to minutes to all structures in the 95% sequence identity filtered non-redundant PDB. A web server front-end for LabelHash as well as a command line version are available at http://labelhash.kavrakilab.org [3]. The LabelHash method is sufficiently fast that it can be used to perform a detailed analysis of the structural variability within large protein families or even superfamilies. Structural variations caused by a wide range of physicochemical and biological sources directly influence the function of a protein. Comparative analysis of drug-receptor substructures across and within species has been used for lead evaluation. Substructurelevel similarity between the binding sites of functionally similar proteins has also been used to identify instances of convergent evolution among proteins. The Family-wise Analysis of SubStructural Templates (FASST) method uses LabelHash for allagainst-all substructure comparison to determine substructural clusters [1]. Substructural clusters characterize the binding site substructural variation within a protein family. We focus on examples of automatically determined substructural clusters that can be linked to phylogenetic distance between family members (see Figure 2), segregation by conformation, and organization by homology among convergent protein lineages. The Motif Ensemble Statistical Hypothesis (MESH) framework constructs a representative motif for each protein cluster among the substructural clusters determined by FASST to build motif Fig. 1. A substructure match (in green) shown superimposed with a motif (in white), while the rest of the matching protein is shown in ribbon representation. Figure reproduced from [2].