A hub-attachment based method to detect functional modules from confidence-scored protein interactions and expression profiles

Abstract

BackgroundMany research results show that the biological systems are composed of functional modules. Members in the same module usually have common functions. This is useful information to understand how biological systems work. Therefore, detecting functional modules is an important research topic in the post-genome era. One of functional module detecting methods is to find dense regions in Protein-Protein Interaction (PPI) networks. Most of current methods neglect confidence-scores of interactions, and pay little attention on using gene expression data to improve their results.ResultsIn this paper, we propose a novel hub-attachment based method to detect functional modules from confidence-scored protein interactions and expression profiles, and we name it HUNTER. Our method not only can extract functional modules from a weighted PPI network, but also use gene expression data as optional input to increase the quality of outcomes. Using HUNTER on yeast data, we found it can discover more novel components related with RNA polymerase complex than those existed methods from yeast interactome. And these new components show the close relationship with polymerase after functional analysis on Gene Ontology.ConclusionA C++ implementation of our prediction method, dataset and supplementary material are available at http://hub.iis.sinica.edu.tw/Hunter/. Our proposed HUNTER method has been applied on yeast data, and the empirical results show that our method can accurately identify functional modules. Such useful application derived from our algorithm can reconstruct the biological machinery, identify undiscovered components and decipher common sub-modules inside these complexes like RNA polymerases I, II, III.

Spotlight: Assembly of protein complexes by integrating graph clustering methods

“…As a parameter-free clustering method (Chin et al, 2010), HUNTER generates a module seed from a vertex and, then, the seed grows gradually by adding vertices that are strongly connected to it. The method further merges any two grown modules with common vertices above a threshold iteratively, and finally determines the output clusters.…”

“…Among the various community structure detection (or graph clustering) methods applied to the PPI network to detect protein complexes and functional modules include random walk based methods (Enright et al, 2002;Pons and Latapy, 2005;van Dongen, 2000), edge betweenness-based methods (Dunn et al, 2005;Girvan and Newman, 2002;Luo et al, 2007), clique percolation methods (Adamcsek et al, 2006;Zhang et al, 2006), and core-attachment based methods (Chin et al, 2010;Leung et al, 2009;Liu et al, 2009;Wu et al, 2009). While relying on widely divergent approaches, these methods have their own unique strengths and limitations.…”

“…the interactions) if the input PPIs are described by standard UniProt ID or yeast SGD IDs. For user convenience, the proposed approach includes other conventionally adopted network clustering methods (e.g., CPM (Palla et al, 2005), FastGreedy (Clauset et al, 2004), HUNTER (Chin et al, 2010), LE (Newman, 2006), MCL (van Dongen, 2000), SpinGlass (Reichardt and Bornholdt, 2006), and WalkTrap (Pons and Latapy, 2005)) in Spotlight platform to easily perform network analysis, as well as view/export, and link results to further functional analysis processes. The Spotlight-based integrating graph clustering method outperforms other network clustering methods by exploiting unique network properties that imply the functional association among proteins.…”

Protein annotation from protein interaction networks and Gene Ontology

“…The MRF methods were later extended by combining PPI data, with gene expression data, protein motif information, mutant phenotype data, and protein localization data to specify which proteins might be active in a given biological process [36,37]. Other global approaches integrated PPI network with more heterogeneous data sources (such as large-scale two-hybrid screens and multiple microarray analyses) [10,38]. Our algorithm ClusFCM [39] assigned biological homology scores to interacting proteins and performed agglomerative clustering on the weighted network to cluster the proteins by known functions and cellular location; functions then were assigned to proteins by a Fuzzy Cognitive Map.…”

A Novel Protein Complex Identification Algorithm Based on the Integration of Local Network Topology and Gene Ontology

“…Many researchers applied this concept to design detecting protein complex, such as COACH [8], Hunter [9], and Core [10].…”