Abstract
We present a new database and an on-line search engine, which store and query the protein binding pockets that interact with single-stranded RNA nucleotide bases. The database consists of a classification of binding sites derived from protein-RNA complexes. Each binding site is assigned to a cluster of similar binding sites in other protein-RNA complexes. Cluster members share similar spatial arrangements of physico-chemical properties, thus can reveal novel similarity between proteins and RNAs with different sequences and folds. The clusters provide 3D consensus binding patterns important for protein-nucleotide recognition. The database search engine allows two types of useful queries: first, given a PDB code of a protein-RNA complex, RsiteDB can detail and classify the properties of the protein binding pockets accommodating extruded RNA nucleotides not involved in local RNA base pairing. Second, given an unbound protein structure, RsiteDB can perform an on-line structural search against the constructed database of 3D consensus binding patterns. Regions similar to known patterns are predicted to serve as binding sites. Alignment of the query to these patterns with their corresponding RNA nucleotides allows making unique predictions of the protein-RNA interactions at the atomic level of detail. This database is accessible at http://bioinfo3d.cs.tau.ac.il/RsiteDB.