Graphs provide a natural framework for visualizing and analyzing networks of many types, including biological networks. Network clustering is a valuable approach for summarizing the structure in large networks, for predicting unobserved interactions, and for predicting functional annotations. Many current clustering algorithms suffer from a common set of limitations: poor resolution of top-level clusters; over-splitting of bottom-level clusters; requirements to pre-define the number of clusters prior to analysis; and an inability to jointly cluster over multiple interaction types. Results A new algorithm, Hierarchical Agglomerative Clustering (HAC), is developed for fast clustering of heterogeneous interaction networks. This algorithm uses maximum likelihood to drive the inference of a hierarchical stochastic block model for network structure. Bayesian model selection provides a principled method for collapsing the fine-structure within the smallest groups, and for identifying the top-level groups within a network. Model scores are additive over independent interaction types, providing a direct route for simultaneous analysis of multiple interaction types. In addition to inferring network structure, this algorithm generates link predictions that with cross-validation provide a quantitative assessment of performance for real-world examples. Conclusions When applied to genome-scale data sets representing several organisms and interaction types, HAC provides the overall best performance in link prediction when compared with other clustering methods and with model-free graph diffusion kernels. Investigation of performance on genome-scale yeast protein interactions reveals roughly 100 top-level clusters, with a long-tailed distribution of cluster sizes. These are in turn partitioned into 1000 fine-level clusters containing 5 proteins on average, again with a long-tailed size distribution. Top-level clusters correspond to broad biological processes, whereas fine-level clusters correspond to discrete complexes. Surprisingly, link prediction based on joint clustering of physical and genetic interactions performs worse than predictions based on individual data sets, suggesting a lack of synergy in current high-throughput data.
Park and BaderBMC Bioinformatics2011,12(Suppl 1):S44 http://www.biomedcentral.com/14712105/12/S1/S44
R E S E A R C HOpen Access Resolving the structure of interactomes with hierarchical agglomerative clustering 1,2 1,2* Yongjin Park, Joel S Bader FromThe Ninth Asia Pacific Bioinformatics Conference (APBC 2011) Inchon, Korea. 1114 January 2011
Abstract Background:Graphs provide a natural framework for visualizing and analyzing networks of many types, including biological networks. Network clustering is a valuable approach for summarizing the structure in large networks, for predicting unobserved interactions, and for predicting functional annotations. Many current clustering algorithms suffer from a common set of limitations: poor resolution of toplevel clusters; oversplitting of bottomlevel clusters; requirements to predefine the number of clusters prior to analysis; and an inability to jointly cluster over multiple interaction types. Results:A new algorithm, Hierarchical Agglomerative Clustering (HAC), is developed for fast clustering of heterogeneous interaction networks. This algorithm uses maximum likelihood to drive the inference of a hierarchical stochastic block model for network structure. Bayesian model selection provides a principled method for collapsing the finestructure within the smallest groups, and for identifying the toplevel groups within a network. Model scores are additive over independent interaction types, providing a direct route for simultaneous analysis of multiple interaction types. In addition to inferring network structure, this algorithm generates link predictions that with crossvalidation provide a quantitative assessment of performance for realworld examples. Conclusions:When applied to genomescale data sets representing several organisms and interaction types, HAC provides the overall best performance in link prediction when compared with other clustering methods and with modelfree graph diffusion kernels. Investigation of performance on genomescale yeast protein interactions reveals roughly 100 toplevel clusters, with a longtailed distribution of cluster sizes. These are in turn partitioned into 1000 finelevel clusters containing 5 proteins on average, again with a longtailed size distribution. Toplevel clusters correspond to broad biological processes, whereas finelevel clusters correspond to discrete complexes. Surprisingly, link prediction based on joint clustering of physical and genetic interactions performs worse than predictions based on individual data sets, suggesting a lack of synergy in current highthroughput data.
Background Graphs or networks provide an excellent organizing fra mework for representing data from highthroughput experiments that measure interactomes, or genome scale biological interactions: physical interactions between proteins; genetic interactions or specific pheno types such as synthetic lethality between genes; gene regulation interactions between transcription factors and
* Correspondence: joel.bader@jhu.edu 1 Department of Biomedical Engineering, Johns Hopkins University, Baltimore, MD 21218, USA Full list of author information is available at the end of the article
genes; and metabolic connections between enzymes and metabolites. In these networks, vertices represent genes, proteins, or other molecules, and edges represent speci fic interaction types [1,2]. An important current challenge is to develop methods to analyze these and other networks, such as social net works [3]. One challenge is to infer network structure by identifying subgroups of related vertices, which in the biological domain may be inferred to have similar func tions. A second challenge is to predict links that might exist but which are not represented in the data. Missing links are prevalent in biological interactomes, where