Locality-sensitive hashing and biological network alignment

Large biological networks contain much information about the functionality of protein-protein interactions and other macromolecular relationships. There are certain small subsets of these graphs which display similar characteristics that appear more frequently than others; these motifs can be used to uncover key information about the structure of the networks. Existing algorithms use exact identity to determine motifs. Here we instead use locality-sensitive hashing to detect networks which have similar but not necessarily identical topologies, and present results for several higher-order motifs. See http://pages.cs.wisc.edu/∼legault for source code.