Skip to main content

Clique-based data mining for related genes in a biomedical database



Progress in the life sciences cannot be made without integrating biomedical knowledge on numerous genes in order to help formulate hypotheses on the genetic mechanisms behind various biological phenomena, including diseases. There is thus a strong need for a way to automatically and comprehensively search from biomedical databases for related genes, such as genes in the same families and genes encoding components of the same pathways. Here we address the extraction of related genes by searching for densely-connected subgraphs, which are modeled as cliques, in a biomedical relational graph.


We constructed a graph whose nodes were gene or disease pages, and edges were the hyperlink connections between those pages in the Online Mendelian Inheritance in Man (OMIM) database. We obtained over 20,000 sets of related genes (called 'gene modules') by enumerating cliques computationally. The modules included genes in the same family, genes for proteins that form a complex, and genes for components of the same signaling pathway. The results of experiments using 'metabolic syndrome'-related gene modules show that the gene modules can be used to get a coherent holistic picture helpful for interpreting relations among genes.


We presented a data mining approach extracting related genes by enumerating cliques. The extracted gene sets provide a holistic picture useful for comprehending complex disease mechanisms.


Progress in the life sciences has recently been made by integrating biomedical knowledge on numerous genes and formulating hypotheses on the genetic mechanisms underlying various vital phenomena [1, 2]. A large variety of genetic and biomedical knowledge on genes has been compiled into databases [3], and is available in electronic forms such as the Online Mendelian Inheritance in Man (OMIM) database [4]. Researchers and physicians formulating hypotheses often need to identify groups of functionally related genes, such as gene families and gene pathways, and this is usually done by simply reading a large number of documents related to the phenomenon of interest [5]. Since such an approach will inevitably result in some relevant literature being overlooked, researchers and physicians need a way that will help them search for related gene sets automatically and comprehensively [6].

Graph-based approaches [79] have recently emerged as a method for data mining. A biomedical relational graph is formed by nodes that represent biological entities (e.g. genes/proteins) and edges that represent the associations of those entities. For instance, protein-protein interactions are modeled by a graph, where nodes are proteins and two nodes are connected by an edge if the corresponding proteins physically bind. Protein functions are predicted using connections in a graph [10] based on the assumption that proteins which lie close to one another are more likely to have similar functions or constitute protein complexes. For extracting coherent groups of genes as modules, a module-assisted approach [11, 12] has been introduced. Prior studies include attempts to extract modules from protein-protein interactions [13, 14], co-expression in microarray data [15, 16], and gene symbol co-occurrence in Medline article abstracts [17]. Computational tools for visualizing modules (sets of nodes) in a given graph [18] have been developed.

Here we have constructed a biomedical relational graph whose nodes are pages of genes or diseases and whose edges are hyperlink connections between pages by using over 10,000 entities in the OMIM database [4]. The OMIM database, which is a biomedical database of human genes and genetic disorders, contains a great number of relationships between genes and diseases. This work was based on the assumption that the structures of hyperlink connections correspond to the structural features of biological systems. Clique-based data mining has been applied to a relational graph based on the assumption that relevant relationships are reflected in completely interconnected subgraphs (cliques) or nearly completely interconnected subgraphs (pseudo-cliques). We address the extraction of related genes (called 'gene modules' in this paper) by searching for densely connected subgraphs in a biomedical relational graph. Sets of related genes are detected by enumerating densely-connected subgraphs modeled as cliques [1921] or pseudo-cliques [22, 23]. Using this method, we extracted over 20,000 gene modules. To the best of our knowledge, this is the first study to show that sets of related genes can be comprehensively extracted from a biomedical database and that these related genes can be utilized to gain insight into the mechanisms of complex diseases.



The experimental materials were taken from the Online Mendelian Inheritance in Man (OMIM) database, a well-known catalog in which human genes and genetic disorders are assigned descriptive code numbers [24]. Using these numbers, they are connected by hyperlinks according to their associations, such as physical proximity, similarity of nomencalture or structure, or functional association.

KEGG pathway data that contains sets of genes were obtained (November 2003) from a database produced by the Cancer Genome Anatomy Project [25].

Clique enumeration

An undirected graph G consists of a set V of nodes and a set E of non-weighted edges connecting pairs of nodes. The number of edges connected to node v V is referred to as the degree of node v in G. The subgraph of G is induced by the subset V' of V. A subgraph in which every pair of nodes is connected by an edge (i.e., a complete subgraph) is called a clique, and the size of which is the number of nodes in it. A clique is called maximal if it is included in no other clique. Pseudo-cliques are the subgraphs obtained by relaxing the connectivities, and the connectivity is measured by the edge density ρ. The edge density of a pseudo-clique S is the ratio of the number of edges in S to the number of edges in a clique that has the same number of nodes that S does. It is calculated as follows:

where |E(S)| is the number of edges in S and |S| is the number of nodes in S. By setting ρ(S) to a threshold value θ (0 ≤ θ ≤ 1) pseudo-cliques whose edge density is not less than θ can be enumerated [26]. As a θ decreases from 1 to 0, we get pseudo-cliques whose connectivity is more and more relaxed.

In the work presented in this paper we obtained gene modules (sets of biologically related genes) by enumerating pseudo-cliques in a graph whose node were genes and genetic disorders and whose connecting edges were biomedical relations.

Correspondence analysis

Correspondence analysis [27] is a method to analyze relations between categorical variables, called cases and items. This analysis yields an arrangement in which similar cases and items are closely placed. By introducing a data matrix whose rows and columns are variables (cases and items) having values of 0 or 1 at components depending on the absence or presence of the relations between the variables, rows (items) and columns (cases) are arranged by sorting the scores calculated using the second largest eigenvalue and the corresponding eigenvector.

In the work presented in this paper items and cases are assigned respectively to modules and genes/diseases contained.

Results and discussion

An example: enumerating cliques

We will present an example of enumerating cliques from a part of an undirected graph used in experiments described later in this paper. This enables us to see clique enumeration from a biomedical relational graph and to investigate the validity in an accessible fashion.

Figure 1 shows a biomedical relational graph in which hypertension and 16 hypertension-related genes are represented by nodes and in which the biomedical associations between pairs of genes and between genes and hypertention are represented by edges. The graph has 17 nodes and 31 edges. It is understandable that a computational device is required in order to grasp the structure of relationship among genes and diseases as the scale becomes large.

Figure 1

Example of a biomedical relational graph. Hypertension and hypertension-related genes are represented by nodes, and the associations between them are represented by edges.

The node sets for all the maximal cliques obtained from the graph in Figure 1 are listed in Table 1. It is expected that there are many cliques that contain the hypertension node. The extraction of cliques whose nodes include hypertension and the AGT, ACE, AGTR1, and REN genes, is in agreement with the medical knowledge. The renin-angiotensin system in which the REN, AGT, ACE, and AGTR1 genes interact with each other [28] is the well known pressor mechanism that acts in concert with the CYP11B1 and CYP11B2 genes related to aldosterone secretion. The extraction of cliques whose nodes include the CYP17 and HSD11B2 genes in addition to the CYP11B1 and CYP11B2 genes is consistent with the reported interaction of their enzymes with aldosterone synthesis [29]. Related genes that regulate each other can thus be extracted by enumerating cliques in biomedical relational graphs.

Table 1 All maximal cliques obtained from the graph in Figure 1

Biomedical relational graph by OMIM hyperlinks

We constructed an undirected graph whose nodes were gene or disease pages in the OMIM database and whose edges were hyperlink connections between those pages. As shown by the data listed in Table 2, the graph had 13,722 nodes and 35,749 edges (as of December 2001). Each hyperlink connection was counted only once. We limited this study to 6,010 genes both in the OMIM database and in Swiss-Prot [30] (November 2002), excluding genes with no Medline citations or no gene locus descriptions or no hyperlinks to other OMIM pages. The characteristic path length is defined as the number of edges in the shortest path between two nodes, averaged over all pairs of nodes. The clustering coefficient measures the average degree of node coherence connected by the edges. When a node V has k v connected nodes, the clustering coefficient is the ratio of the actual number of edges to the possible k v (k v -1)/2 edges. The characteristic path length and clustering coefficient [31] were respectively 4.99 and 0.27 (Table 2). These values indicate that the graph contained clusters of densely-connected nodes and that there were hub-like nodes connecting the clusters.

Table 2 Structural properties of the graph used in the experiment

Extraction of gene modules by searching cliques

In the graph were 20,486 maximal cliques and the largest maximal clique contained 12 nodes. The 20 most frequent genes in the maximal cliques are listed in Table 3 along with their degrees and the number of times they were found in cliques of various sizes. The NFKB1 gene (the 3rd most frequent gene, or #3) was also in five cliques of size 10, the BRCA1 gene (#11) in one clique of size 12, and the TAF1 gene (#19) in one clique of size 8. The genes found most frequently in the maximal cliques (such as TNF, TP53, and NFKB1) are typical genes that play a central role and are prevalent research subjects. The clique distribution of the TAF1 gene (#19) is relatively shifted towards larger sizes, suggesting it forms complexes with some gene products since the gene function is a transcription factor.

Table 3 The 20 most frequent genes in the maximal cliques.

When pseudo-cliques were extracted by the relaxing edge connecting condition, the maximum sizes increased as θ decreased (Table 4). Preliminary experimental results showed that gene modules should consist of possible related genes for the analysis. By relaxing connectivities (see Methods), pseudo-cliques were introduced for gene modules to collect the possible related genes. The maximum pseudo-clique size reached 14 when θ = 0.88 and did not increase further even when θ was decreased to 0.7 (data not shown).

Table 4 Number of cliques corresponding to θ values

For this reason, 25,642 pseudo-cliques (θ = 0.88) were taken as gene modules in the following analysis. The sensitivity and specificity of the current method in extracting biologically relevant genes using cliques can not be readily assessed. One would assume that cliques with larger size are likely to be more biologically relevant. To account for this, we estimated whether a cutoff value on the clique size can be set. We examined the maximum size of cliques in randomized graphs having common node degrees to the OMIM hyperlink graph. In the experiments of enumerating pseudo-cliques (θ = 0.88) in the randomized graphs which were generated using an edge-swapping approach [32], the maximum size four was observed (data not shown). This result suggests that clique size four may be considered as a background clique size, and that clique size of more than four may be biologically relevant.

While we employed the clique enumeration method to extract sets of related genes, application of the method by the edge-betweenness clustering [11, 12], which introduces the concept of graph modularity called community, may also be considered. Basically the community-based approach is conducted with graph partitioning to separate connected nodes into groups of nodes that have a high density of edges within them, with a lower density of edges between groups [13]. In contrast, the clique-based approach allows for any node to belong to more than one group of nodes. It would be interesting to compare the performance of these different methods and investigate how to use the methods properly.

Biological evaluation of extracted gene modules

To evaluate how well the gene modules correspond to known gene pathways, we compared the sets of genes in the modules with those in the KEGG pathway database [33]. The 66 KEGG pathways that contain more than five genes and the 25,642 gene modules (θ = 0.88) were compared by calculating values of the Jaccard's coefficient r J (0 ≤ r J ≤ 1). The Jaccard's coefficient is often used as a criterion when evaluating the similarity of two sets, and is the ratio of the size of the intersection of the sets to the size of the union of the sets. Its value thus approaches 1 as the extent of coincidence increases. The 15 most relevant KEGG pathways and the corresponding gene module sizes are listed in Table 5, where the numbers in the square brackets are numbers of genes in the KEGG pathway. For instance, in the top rank of KEGG pathway 'Blood Group Glycolipid Biosynthesis,' the number of genes contained was six. The corresponding gene module by a pseudo-clique included four genes. As for the other KEGG pathways, none of the gene module sizes equaled the number of genes in the KEGG pathway. This implies that a pseudo-clique can partially extract genes included in biological pathways.

Table 5 The 15 KEGG pathways most relevant to the extracted gene modules

In order to evaluate how well the gene modules accord with interactions between genes, we compared the sets of genes in the modules with gene pairs in the protein-protein interaction data [34]. The 650 interactions for the above-mentioned 6,010 genes were obtained from the protein-protein interaction data. Among the 650 gene pairs 145 pairs were captured in the gene modules, which correspond to 22.3% of the total interactions. The gene pairs in the protein-protein interactions which were repeatedly captured by gene modules are listed in Table 6. The gene pair in the first row is involved in genes of biological pathways related to MAPK signaling and apoptosis. The second pair is involved in TGF-β signaling. The third is in regulation of food intake and energy monitoring. Repeated captures suggest that the pairs have various functions in biological setting.

Table 6 The captured gene pairs in the protein-protein interaction database

Table 7 shows gene sets included in typical large gene modules. The gene module in the first row is constituted by a family of chemokine genes, and the gene module in the second comprises NF-κ B family genes (including RelA and RelB) and genes that form complexes with them (Iκ B). The gene module in the third row is made up of 'DNA repair'-related genes. The BRCA1-associated proteins; the BLM, MSH6, MSH2, and MLH1 proteins; and subunits of the RFC complex are involved in DNA repair [35]. The genes in the module in the fourth row are related to general transcription factor (GTF) protein complexes. The gene module in the bottom row is associated with the signal transduction pathway of the inflammatory response [36]. TNF receptor-associated factor 2 (TRAF2) is a protein that interacts with TNF receptors and is required for signal transduction. The MAP kinase kinase kinase 14 (MAP3K14) gene in this module encodes a protein that simulates NF-κ B activity by binding to the TRAF2 gene product. The gene modules thus comprise various types of related genes including gene families, complexes, and pathways. As illustrated above, the current results have biological coherence for analyzing relations among genes.

Table 7 Typical large gene modules computationally extracted as pseudo-cliques

Analysis of gene relationships using 'metabolic syndrome'-related gene modules

For applying gene modules to disease mechanism analysis, we assembled gene modules associated with the metabolic syndrome [37] as an example of a typical multifactorial disease. The metabolic syndrome is a heterogeneous disease characterized by the onset and progession of four common disorders: obesity, diabetes, hyperlipidemia, and hypertension. The genes associated with these disorders might interact with each other and lead to arteriosclerotic diseases such as myocardial infarction or ischemic stroke. Public attention has been focused on its prevention [38]. We examined the congruence with current medical knowledge.

Gene modules associated with diabetes, hyperlipidemia, hypertension, and obesity are obtained by their containing the disease nodes, which are non-insulin-dependent diabetes mellitus (MIM 125853), familial combined hyperlipidemia (144250), essential hypertension (145500), and obesity (601665). Out of 25,642 modules, 110, 16, 34, and 28 modules are obtained, respectively. There were no overlaps among the modules. Then a total of 188 modules and 124 genes contained were identified.

The 10 most frequent genes in the 188 modules are listed in Table 8 along with the numbers of times they were found in the modules (i.e., cliques) of various sizes. As shown in the table, INS gene and LEP gene are the top and the 2nd, respectively. The modules of size 6 including INS gene or LEP gene were {Obesity, LEP, MC4R, POMC, AGRP, LEPR}, {Obesity, LEP, MC4R, POMC, AGRP, PCSK1}, {Diabetes, LEP, IGF1, IRS1, INS, IRS2}. Each module contains biologically plausible genes related to obesity or diabetes.

Table 8 The 10 most frequent genes in the 188 extracted modules associated with the metabolic syndrome

We combined the 188 modules and 124 genes using correspondence analysis (Fig. 2). Each bar in Figure 2 corresponds to a gene node or disease node in the cliques, and the names of the genes or diseases are shown on the right. Each bar is colored according to the disease nodes included in the corresponding module: red for hypertension, gray for hyperlipidemia, blue for obesity, and green for diabetes. The letters of 't,' 'l,' 'o,' and 'd' to the right of the figure respectively indicate that in the literature the genes in the corresponding rows are related to hypertension [39], hypertrigricedemia [40], obesity [41], and diabetes [42].

Figure 2

188 modules associated with obesity, diabetes, hyperlipidemia, and hypertension. The vertical bars indicate the genes/diseases in the modules, the columns represent modules, and the rows represent the genes/diseases. The rows and columns are sorted in the ascending order by the score calculated by correspondence analysis (see Methods). Red, grey, blue, and green bars respectively indicate cliques that contain the hypertension, hyperlipidemia, obesity, and diabetes nodes. The letters 't,' 'l,' 'o,' and 'd' on the right show that in the literature the genes are related to hypertension, hyperlipidemia, obesity, and diabetes (see text for the literature references).

Hyperlipidemia-related genes are replaced by genes associated with hypertrigricedemia (hyperlipidemia together with hypercholesterolemia). As shown in Figure 2, the gene modules associated with the four diseases occupy their own regions from the top-left to the bottom-right. As for the disease nodes shown to be belt-like in the figure, the nodes of hypertension, hyperlipidemia, obesity, diabetes are placed from the top to the middle. There is a diabetes node at the middle and the letter 'd' indicating diabetes-related genes are mostly clustered in the lower half. This implies that gene modules properly reflect the biological functions of the corresponding genes and that the modules can provide a holistic view of a complex disease. Correspondence analysis has been successfully applied to combine gene modules for interpreting relationships among genes with their gradation on relation to diseases.

Genes related to the above-mentioned aldosterone metabolism (CYP17, HSD11B2, CYP11B1, CYP11B2) are at the top of the figure. Genes related to the renin-angiotensin system (REN, AGT, AGTR1, ACE) are around the hypertension node indicated by the cluster of 't' marks, and apolipoprotein genes (APOA2, APOC3, APOE, APOA1, APOB) are gathered around the hyperlipidemia node close to the 'renin-angiotensin system'-related genes. It is worth noting that 'food intake regulation'-related genes such as LEP, NPY, MC4R, AGRP and POMC genes are grouped comprehensively (i.e., without overlooking relevant genes). Insulin resistance is defined as a status in which the action of insulin is insufficient and proper energy conversion is impaired. The TNF gene related to insulin resistance [43], which is a shared foundation of the four disorders, was in modules associated with hypertension, obesity, and diabetes, indicating that it has a variety of effects on the development of metabolic syndrome. It would be intriguing to find the CRH gene, which has a biological function in stress responses based on the hypothalamic-pituitary-adrenal axis, in the vicinity of the 'food intake regulation'-related gene region. FGF (fibroblast growth factor)-family genes appear at the bottom of the figure, indicating their relevance to diabetes. Although FGF-genes are associated with cancers, inflammation has been investigated in both cancer and metabolic syndrome. Such findings in this chart suggest innovative avenues of research. Obtaining the comprehensive list of related genes makes it possible to combine the gene modules and grasp relations among genes quantitatively, facilitating hypothesis formulation from a holistic viewpoint. In principle, the biological functions of desired genes as well as disease-related genes could be grasped by combining modules that contain those genes.


We have shown that related genes can be extracted comprehensively by enumerating pseudo-cliques in biomedical relational graph. Over 20,000 gene modules that include genes in the same family, genes encoding proteins in the same complexes, and genes encoding components of the same signaling pathway were extracted automatically. Furthermore these gene modules were utilized for visualizing relations between genes and diseases. Extraction of related genes (referred to as gene modules in this paper) would be more important since most biomedical tasks are performed not by individual genes, but by sets of functionally associated genes. The method using pseudo-cliques can generally extract related gene sets under a single computational operation although the application is restricted to genes connected by their relations in the manner of the OMIM hyperlink connections. For clarifying complex disease mechanisms, obtaining relationships among genes associated with the diseases should be crucial. Comprehensive extraction of related-gene sets by clique-based data mining may provide us with a systematic methodology for gaining insight into the genetic mechanisms underlying various biological phenomena including diseases.


  1. 1.

    Jensen LJ, Saric J, Bork P: Literature mining for the biologist: from information retrieval to biological discovery. Nat Rev Genet 2006, 7: 119–129.

    CAS  Article  PubMed  Google Scholar 

  2. 2.

    Matsunaga T, Muramatsu M: Disease-related concept mining by knowledge-based two-dimensional gene mapping. J Bioinform Comput Biol 2007, 5: 1047–1067.

    CAS  Article  PubMed  Google Scholar 

  3. 3.

    Galperin MY: The molecular biology database collection: 2008 update. Nucleic Acids Res 2008, 36: D2-D4.

    PubMed Central  CAS  Article  PubMed  Google Scholar 

  4. 4.

    Hamosh A, Scott AF, Amberger J, Valle D, McKusick VA: Online Mendelian Inheritance in Man (OMIM). Hum Mutat 2000, 15: 57–61.

    CAS  Article  PubMed  Google Scholar 

  5. 5.

    Oda K, Matsuoka Y, Funahashi A, Kitano H: A comprehensive pathway map of epidermal growth factor receptor signaling. Mol Syst Biol 2005., 1: 2005.0010 2005.0010

    Google Scholar 

  6. 6.

    Roberts PM: Mining literature for systems biology. Brief Bioinform 2006, 7: 399–406.

    CAS  Article  PubMed  Google Scholar 

  7. 7.

    Cook DJ, Holder LB: Graph-based data mining. IEEE Intelligent Systems 2000, 15: 32–41.

    Article  Google Scholar 

  8. 8.

    Barabási AL, Oltvai ZN: Network biology: understanding the cell's functional organization. Nat Rev Genet 2004, 5: 101–113.

    Article  PubMed  Google Scholar 

  9. 9.

    Aittokallio T, Schwikowski B: Graph-based methods for analysing networks in cell biology. Brief Bioinform 2006, 7: 243–255.

    CAS  Article  PubMed  Google Scholar 

  10. 10.

    Sharan R, Ulitsky I, Shamir R: Network-based prediction of protein function. Mol Syst Biol 2007, 3: 88.

    PubMed Central  Article  PubMed  Google Scholar 

  11. 11.

    Newman MEJ, Girvan M: Finding and evaluating community structure in networks. Phys Rev 2004, E69: 026113.

    Google Scholar 

  12. 12.

    Newman MEJ: Detecting community structure in networks. Eur Phys J 2004, B38: 321–330.

    Article  Google Scholar 

  13. 13.

    Dunn R, Dudbridge F, Sanderson CM: The use of edge-betweenness clustering to investigate biological function in protein interaction networks. BMC Bioinformatics 2005, 6: 39.

    PubMed Central  Article  PubMed  Google Scholar 

  14. 14.

    Chen J, Yuan B: Detecting functional modules in the yeast protein-protein interaction network. Bioinformatics 2006, 22: 2283–2290.

    CAS  Article  PubMed  Google Scholar 

  15. 15.

    Hu H, Yan X, Huang Y, Han J, Zhou XJ: Mining coherent dense subgraphs across massive biological networks for functional discovery. Bioinformatics 2005, 21: i213-i221.

    CAS  Article  PubMed  Google Scholar 

  16. 16.

    Yan X, Mehan MR, Huang Y, Waterman MS, Yu PS, Zhou XJ: A graph-based approach to systematically reconstruct human transcriptional regulatory modules. Bioinformatics 2007, 23: i577-i586.

    CAS  Article  PubMed  Google Scholar 

  17. 17.

    Wilkinson DM, Huberman BA: A method for finding communities of related genes. Proc Natl Acad Sci USA 2004, 101: 5241–5248.

    PubMed Central  CAS  Article  PubMed  Google Scholar 

  18. 18.

    Adamcsek B, Palla G, Farkas IJ, Dereńyi I, Vicsek T: CFinder: locating cliques and overlapping modules in biological networks. Bioinformatics 2006, 22: 1021–1023.

    CAS  Article  PubMed  Google Scholar 

  19. 19.

    Zhang Y, Abu-Khzam FN, Baldwin NE, Chesler EJ, Langston MA, Samatova NF: Genome-scale computational approaches to memory-intensive applications in systems biology. Proc ACM/IEEE Conf Supercomputing 2005, 12.

    Google Scholar 

  20. 20.

    Tomita E, Tanaka A, Takahashi H: The worst-case time complexity for generating all maximal cliques and computational experiments"(Invited paper for the special issue on COCOON 2004). Teoret Comput Sci 2006, 363: 28–42.

    Article  Google Scholar 

  21. 21.

    Tomita E: The maximum clique problem and its applications-invited lecture-. IPSJ SIG Technical Report 2007, 21–24.

    Google Scholar 

  22. 22.

    Haraguchi M, Okubo Y: A method for pinpoint clustering of web pages with pseudo-clique search. Lecture Notes in Artificial Intelligence 2006, 3847: 59–78.

    Google Scholar 

  23. 23.

    Uno T: An efficient algorithm for enumerating pseudo cliques. Lecture Notes in Computer Science 2007, 4835: 402–414.

    Article  Google Scholar 

  24. 24.

    OMIM text (omim.txt)[]

  25. 25.

    Gene annotations (Hs_GeneData.dat)[]

  26. 26.

    Clique/pseudo-clique enumeration program (PCE)[]

  27. 27.

    Benzecri JP: Correspondense analysis handbook. New York: Marcel Dekker; 1992.

    Google Scholar 

  28. 28.

    Baudin B: Angiotensin II receptor polymorphisms in hypertension. Pharmacogenomic considerations. Pharmacogenomics 2002, 3: 65–73.

    CAS  Article  PubMed  Google Scholar 

  29. 29.

    White PC, Agarwal AK, Nunez BS, Giacchetti G, Mantero F, Stewart PM: Genotype-phenotype correlations of mutations and polymorphisms in HSD11B2, the gene encoding the kidney isozyme of 11beta-hydroxysteroid dehydrogenase. Endocr Res 2000, 26: 771–780.

    CAS  Article  PubMed  Google Scholar 

  30. 30.

    Bairoch A, Apweiler R: The SWISS-PROT protein sequence database: its relevance to human molecular medical research. J Mol Med 1997, 75: 312–316.

    CAS  PubMed  Google Scholar 

  31. 31.

    Watts DJ, Strogatz SH: Collective dynamics of 'small-world' networks. Nature 1998, 393: 440–442.

    CAS  Article  PubMed  Google Scholar 

  32. 32.

    Müller H, Mancuso F: Idetification and analysis of co-occurrence networks with NetCutter. PLoS ONE 2008, 3: e3178.

    PubMed Central  Article  PubMed  Google Scholar 

  33. 33.

    Kanehisa M, Goto S: KEGG: kyoto encyclopedia of genes and genomes. Nucleic Acids Res 2000, 28: 27–30.

    PubMed Central  CAS  Article  PubMed  Google Scholar 

  34. 34.

    Goh KI, Cusick ME, Valle D, Childs B, Vidal M, Barabási AL: The human disease network. Proc Natl Acad Sci USA 2007, 104: 8685–8690.

    PubMed Central  CAS  Article  PubMed  Google Scholar 

  35. 35.

    Wang Y, Cortez D, Yazdi P, Neff N, Elledge SJ, Qin J: BASC, a super complex of BRCA1-associated proteins involved in the recognition and repair of aberrant DNA structures. Genes Dev 2000, 14: 927–939.

    PubMed Central  CAS  PubMed  Google Scholar 

  36. 36.

    Hauer J, Püschner S, Ramakrishnan P, Simon U, Bongers M, Federle C, Engelmann H: TNF receptor (TNFR)-associated factor (TRAF) 3 serves as an inhibitor of TRAF2/5-mediated activation of the noncanonical NF- κ B pathway by TRAF-binding TNFRs. Proc Natl Acad Sci USA 2005, 102: 2874–2879.

    PubMed Central  CAS  Article  PubMed  Google Scholar 

  37. 37.

    NCEP: Executive Summary of The Third Report of The National Cholesterol Education Program (NCEP) Expert Panel on Detection, Evaluation, And Treatment of High Blood Cholesterol In Adults (Adult Treatment Panel III). JAMA 2001, 285: 2486–2497.

    Article  Google Scholar 

  38. 38.

    Eckel RH: Mechanisms of the components of the metabolic syndrome that predispose to diabetes and atherosclerotic CVD. Proc Nutrition Society 2007, 82–95.

    Google Scholar 

  39. 39.

    Halushka MK, Mathews DJ, Bailey JA, Chakravarti A: GIST: A web tool for collecting gene information. Physiol Genomics 1999, 1: 75–81.

    CAS  PubMed  Google Scholar 

  40. 40.

    Seda O: Comparative gene map of hypertriglyceridaemia. Folia Biol 2004, 50: 43–57.

    CAS  Google Scholar 

  41. 41.

    Snyder EE, Walts B, Peŕusse L, Chagnon YC, Weisnagel SJ, Rankinen T, Bouchard C: The human obesity gene map: the 2003 update. Obes Res 2004, 12: 369–439.

    CAS  Article  PubMed  Google Scholar 

  42. 42.

    Almind K, Doria A, Kahn CR: Putting the genes for type II diabetes on the map. Nat Med 2001, 7: 277–279.

    CAS  Article  PubMed  Google Scholar 

  43. 43.

    De Fronzo RA, Ferrannini E: Insulin resistance syndrome. A multifaceted syndrome responsible for NIDDM, obesity, hypertension, dyslipidemia, and atherosclerotic cardiovascular disease. Diabetes Care 1991, 14: 173–194.

    CAS  Article  Google Scholar 

Download references

Author information



Corresponding author

Correspondence to Tsutomu Matsunaga.

Additional information

Authors' contributions

TM and ET conceived and designed the experiments. TM and CY performed the experiments. TM and MM analyzed the data. All authors have read and approved the final manuscript.

Authors’ original submitted files for images

Below are the links to the authors’ original submitted files for images.

Authors’ original file for figure 1

Authors’ original file for figure 2

Rights and permissions

This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Reprints and Permissions

About this article

Cite this article

Matsunaga, T., Yonemori, C., Tomita, E. et al. Clique-based data mining for related genes in a biomedical database. BMC Bioinformatics 10, 205 (2009).

Download citation


  • Gene Module
  • Cluster Coefficient
  • KEGG Pathway
  • Maximal Clique
  • Clique Enumeration