Biological network motif detection and evaluation

Kim et al. BMC Systems Biology 2011, 5(Suppl 3):S5 http://www.biomedcentral.com/1752-0509/5/S3/S5 RESEARCH Open Access Biological network motif detection and evaluation Wooyoung Kim1*, Min Li1,2*, Jianxin Wang2, Yi Pan1* From BIOCOMP 2010 - The 2010 International Conference on Bioinformatics and Computational Biology Las Vegas, NV, USA. 12-15 July 2011 Abstract Background: Molecular level of biological data can be constructed into system level of data as biological networks. Network motifs are defined as over-represented small connected subgraphs in networks and they have been used for many biological applications. Since network motif discovery involves computationally challenging processes, previous algorithms have focused on computational efficiency. However, we believe that the biological quality of network motifs is also very important. Results: We define biological network motifs as biologically significant subgraphs and traditional network motifs are differentiated as structural network motifs in this paper. We develop five algorithms, namely, EDGEGO-BNM, EDGEBETWEENNESS-BNM, NMF-BNM, NMFGO-BNM and VOLTAGE-BNM, for efficient detection of biological network motifs, and introduce several evaluation measures including motifs included in complex, motifs included in functional module and GO term clustering score in this paper. Experimental results show that EDGEGO-BNM and EDGEBETWEENNESS-BNM perform better than existing algorithms and all of our algorithms are applicable to find structural network motifs as well. Conclusion: We provide new approaches to finding network motifs in biological networks. Our algorithms efficiently detect biological network motifs and further improve existing algorithms to find high quality structural network motifs, which would be impossible using existing algorithms. The performances of the algorithms are compared based on our new evaluation measures in biological contexts. We believe that our work gives some guidelines of network motifs research for the biological networks. Background Systems biology focuses on the study of complex interactions in biological systems, rather than the study of individual molecules such as DNA, RNA, proteins and metabolites [1]. One of the goals of systems biology is understanding the structures of all molecules and their interactions in a system level. Therefore major challenges are understanding the dynamic structures of small molecules and determining their functions in a living cell. Various types of biological interactions have been expressed in networks, which include transcriptional regulatory networks, signaling pathways, metabolic networks and protein-protein interaction (PPI) networks. Biological networks share some of structural properties of other complex networks, or have specific features of scale-free * Correspondence: ; ; 1 Department of Computer Science, Georgia State University, Atlanta, USA Full list of author information is available at the end of the article and small-world effect [2]. However, the properties have been questioned by Lacroix et al. [3] with a number of reasons including the incompleteness of networks and inconsistent link generation for the graphs. Therefore, the analysis extends to other network properties such as network clusters and network motifs. As biological networks are massive and the size is still increasing, dividing the network into a number of clusters helps reveal specific local properties. Network motif, as another concept describing local properties of a network, is defined as a small connected subgraph appearing frequently and uniquely in a network. Similar to a protein sequence motif, network motif is defined as a over-repeated pattern, but it requires much more computation as the process involves isomorphic testing and repeated processes for uniqueness determination. Network alignment [4] and network querying [5] are analogous to network motifs, but while network motifs © 2011 Kim et al. This is an open access article distributed under the terms of the Creative Commons Attribution License (http:// creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. Kim et al. BMC Systems Biology 2011, 5(Suppl 3):S5 http://www.biomedcentral.com/1752-0509/5/S3/S5 are defined with only structural information, network alignment and network querying require both of the topological and biological information. Previous network motif discovery algorithms include exact counting and approximation algorithms: Exhaustive recursive search (ERS) [6], enumerate subgraphs (ESU) [7] and compact topological motifs [8] are exact counting algorithms. For efficient detection, several approximation algorithms have been provided including edge sampling (MFINDER) [6], randomized version of ESU from a search tree (RAND-ESU) [9], and tree-filtering search which is NEMOFINDER[10]. Furthermore, parallel search algorithms have been developed to realize feasible exact counting algorithms [11,12]. Network motifs are used for many applications in biological networks. Feed-forward-loop (FFL) and bifan network motifs are identified as the typical patterns in different types of biological networks [13,14]. Przulj et al. [15] used network motifs as a relative graphlet frequency distance to distinguish different protein-protein interaction networks. Also motif frequencies are exploited as classifiers for network model selection [16]. Milo et al. [17] studied that networks of different biological and technological domains have been classified into different superfamilies on the basis of motif significance profiles. To predict protein-protein interactions, Albert I. and Albert R. [18] used network motifs successfully. In the study by Conant and Wagner [19], network motifs in transcriptional regulatory networks are not evolutionary conserved while network motifs in PPI networks are evolutionary related. On the other hand, network motifs are extended to ‘motif modes’ each of which has a certain topology and a specific functional property [20]. Through a number of network motif applications, however, we notice several problems regarding the biological meanings of network motifs, on top of the computational challenge for the detection. First, the biological quality of network motifs are not validated thoroughly. A network motif is selected only by its structural uniqueness and just small number of instances of the type are biologically exemplified. Second, only small portion of network motif instances are used for applications and others are ignored. Third, non-motifs, that is, structurally insignificant subgraphs, have not been analyzed in any studies, which are filtered out before applying to any applications. Fourth, it is still questionable what the network motifs really represent in biological networks. As we believe that the biological quality of network motifs are also significant, we (...truncated)