Genetic interactions (such as synthetic lethal interactions) have become quantifiable on a large-scale using the epistatic miniarray profile (E-MAP) method. An E-MAP allows the construction of a large, weighted network of both aggravating and alleviating genetic interactions between genes. By clustering genes into modules and establishing relationships between those modules, we can discover compensatory pathways. We introduce a general framework for applying greedy clustering heuristics to probabilistic graphs. We use this framework to apply a graph clustering method called graph summarization to an E-MAP that targets yeast chromosome biology. This results in a new method for clustering E-MAP data that we call Expected Graph Compression (EGC). We validate modules and compensatory pathways using enriched Gene Ontology annotations and a novel method based on correlated gene expression. EGC finds a number of modules that are not found by any previous methods to cluster E-MAP data. EGC also uncovers core sub-modules contained within several previously found modules, suggesting that EGC can reveal the finer structure of E-MAP networks. © Springer-Verlag Berlin Heidelberg 2010.
CITATION STYLE
Kelley, D. R., & Kingsford, C. (2010). Extracting between-pathway models from E-MAP interactions using expected graph compression. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 6044 LNBI, pp. 248–262). https://doi.org/10.1007/978-3-642-12683-3_16
Mendeley helps you to discover research relevant for your work.