Distance measures for network comparison
My work involves the development of appropriate topology-based distance measures for large networks and the application of these to biological networks in particular. Due to the extremely noisy nature of experimental molecular interaction data, traditional approaches such as graph matching/network alignment have had limited success with these networks. This is compounded by the highly variable sizes of available datasets across different species. We are thus exploring purely topology based
approaches, which base network similarity upon the presence of small, signature “graphlets” within the larger network. We are developing a method that embeds these graphlet counts within a statistically sound distance measure that could be used to compare arbitrarily complex and variably-sourced networks, even in the complete absence of node-associated information.