Improving Functional Module Detection

There has been a great deal of recent interest in identifying functional modules from protein interaction and gene expression data. One commonly used computational technique is simulated annealing, which while asymptotically correct frequently suffers from slow convergence. In this paper we outline and exploit the analogy between finding functional modules and finding Haplotype Blocks from genetic data, to investigate a new technique for finding functional modules which does not rely on Monte Carlo methodology. We discuss circumstances under which our algorithm may work, but under which simulated annealing may not converge to known modules. We also suggest how our methodology might supplement, and improve the performance, of existing Monte Carlo searches.