- Open Access
Core and periphery structures in protein interaction networks
© Luo et al; licensee BioMed Central Ltd. 2009
- Published: 29 April 2009
Characterizing the structural properties of protein interaction networks will help illuminate the organizational and functional relationships among elements in biological systems.
In this paper, we present a systematic exploration of the core/periphery structures in protein interaction networks (PINs). First, the concepts of cores and peripheries in PINs are defined. Then, computational methods are proposed to identify two types of cores, k-plex cores and star cores, from PINs. Application of these methods to a yeast protein interaction network has identified 110 k-plex cores and 109 star cores. We find that the k-plex cores consist of either "party" proteins, "date" proteins, or both. We also reveal that there are two classes of 1-peripheral proteins: "party" peripheries, which are more likely to be part of protein complex, and "connector" peripheries, which are more likely connected to different proteins or protein complexes. Our results also show that, besides connectivity, other variations in structural properties are related to the variation in biological properties. Furthermore, the negative correlation between evolutionary rate and connectivity are shown toysis. Moreover, the core/periphery structures help to reveal the existence of multiple levels of protein expression dynamics.
Our results show that both the structure and connectivity can be used to characterize topological properties in protein interaction networks, illuminating the functional organization of cellular systems.
- Evolutionary Rate
- Core Protein
- Protein Interaction Network
- Core Member
- Saccharomyces Genome Database
Network biology , which models biological systems as networks of connected elements, enables biologists to understand both macroscopic properties of biological systems [2–5] and microscopic properties of single molecules within systems . With the advances in high-throughput techniques, more and more large-scale biological networks have been defined [7, 8]. Studying the structure of biological networks will help elucidate the organization and functional relationships of elements in cellular systems.
Recently, Guimera et al.  classified the roles of nodes in complex networks according to their properties inside sub-network "modules". Their classification depended on dissecting the network into modules using a simulated annealing method . However, precisely identifying biologically-relevant modules from PINs is not a trivial task. Fortunato and Barthelemy  recently pointed out that the optimization of Newman-Girvan modularity appears to favor large modules, and thus may miss important biological relationships that exist at the molecular level. Application of the method of Guimera and Amaral  to separate the yeast PIN from MIPS  into modules showed that these structurally-defined modules did not show a significant correlation with biological functional units. Thus, defining roles of proteins based on these modules may not be appropriate for PINs. However, it is still possible to understand the roles of proteins in the PINs within other types of sub-graph structures. In this study, we explore the role of proteins in PIN based on core/periphery structures.
Core/periphery structures can be related to protein complexes. Protein complexes often include a static part in which components stably interact with each other all the time and a dynamic part that is assembled in a just-in-time fashion [21, 22]. If the just-in-time assembled proteins only interact with small portion of the static part, the whole protein complex may appear as a core/periphery structure in the PINs. On the other hand, proteins that interact with different proteins in different contexts may emerge as a star structure in the PIN. Thus, the investigation of the core/periphery structure in PINs may help elucidate the dynamic of protein complex.
Furthermore, previous studies have shown that the structural characteristics, like connectivity (number of links), of proteins in PINs is related to the biological properties, such as essentiality  and evolutionary rate [23, 24]. On the other hand, the roles and properties of proteins are also found to be related to the structural characteristics of proteins in the PIN . However, the relationship between structural and biological properties of core and peripheral proteins in PIN has not been fully explored. It is plausible to hypothesize that the core and peripheral proteins may have different roles and properties due to their different topological characteristics. For example, core proteins are usually more highly connected to each other and may have higher essentiality characteristics and lower evolutionary rates than those of peripheral proteins. Combining the structural characteristics of proteins with their biological properties may help elucidate their different roles in biology systems.
In this paper, we present a systematic exploration of core/periphery structures in PINs. Our studies help elucidate the relationship between topological properties in PINs and the roles played by proteins in cellular system, and thus help define the organizational mechanisms used in cellular system.
A PIN can be modeled as an undirected and unweighted graph G = (V, E), where the vertices set V represents proteins and the edges set E represents interactions between proteins. In the context of this paper, the graph is synonymous with the network. A core  in a network is a cohesive sub-graph, in which nodes are highly connected to each other. There are various definitions of cohesive sub-graph based on different connectivity properties of the vertices, including cliques , k-plexes , k-cores  and n-cliques .
A clique is a complete sub-graph of three or more nodes in which all nodes are connected to each other. A maximum clique is a complete sub-graph in the graph such that there are no nodes remaining in the graph that are connected to all the member of the clique. However, the clique is a very restrictive sub-group definition for protein interaction networks. Two concepts in network theory have been proposed to loosen the clique definition. The n-clique relaxes the requirement on the distance between nodes inside the sub-graph. An n-clique is a sub-graph in which all pairs of nodes are no greater than n distance apart. Unfortunately, the n-clique are often not be very cohesive, even for a 2-clique. The k-plex sub-graph definition relaxes the number of nodes required to be connected for each node in the sub-graph. A k-plex is a sub-graph in which each node is connected to at least n-k nodes, where n is the number of nodes in the sub-graph and k is a tunable parameter. Another cohesive sub-graph is the k-core. A k-core is a connected maximal sub-graph in which each node has degree (number of connections) at least k. Although the k-core includes all cohesive sub-graphs, it may also contain non-cohesive parts. Another problem with the k-core approach is that k-cores cannot overlap. Based on these considerations, we expect that the k-plex approach will likely provide a better representation of functionally relevant sub-graph cores in protein interaction networks.
In this study, we define a core in a PIN as a local maximal k-plex with k ≤ n/2 , where n is the number of nodes in the sub-graph. The local maximum means that no more peripheral node can be added into the sub-graph such that the sub-graph remains a k-plex at a given k.
We also define the k-periphery of a core as the set of nodes that are not in the core and whose distances to any member in the core are equal to k. For example, the 1-periphery is the set of nodes that are directly connected to core members (distance equals to 1). Our definition is different from the original definition of k-periphery by Everett and Borgatti , in which the k-periphery also includes nodes whose distances to any member of the core are less than k. Here, we will focus our study on the 1- and 2-peripheries of a core.
One special core/periphery structure is the star. In an ideal star, one single node is the core, and there are no connections between the peripheral nodes. For biological networks, we will allow limited connections between peripheral nodes, which will be controlled by the peripheral degree defined below.
Types of 1-peripheral nodes
Based on how they are connected to the core members, we classify 1-peripheral nodes into the following types: (1) the closed-single-core peripheral nodes (closed), which are only connected to members of one core (node A in Fig. 1); (2) The multiple-core peripheral nodes (multiple-core), which are connected to members of at least two different cores and may also be connected to other non-core nodes (node B in Fig. 1); (3) The open-single-core peripheral nodes (open), which are connected not only to members of one core but to other non-core nodes. This type of peripheral nodes can be further divided into complete-open-single-core peripheral nodes (complete-open), which have fewer connections to core members than to other non-core nodes (node C in Fig. 1), and limited-open-single-core peripheral nodes (limited-open), which have more connections to core members than to other non-core nodes (node D in Fig. 1); (4) The core-member peripheral nodes, which are members of one core and the 1-peripheries of some other cores (node E and F in Fig. 1). The delineation of these 1-peripheral node types will allow us to investigate if these structural distinctions have biological correlates.
Structural measures for 1-peripheries
The characteristics of 1-peripheries can be described by the following structural measures:
Cp of a 1-periphery node is defined as the ratio of the number of its connections to the core members over the total number of core members, 0 < Cp < 1. The coreness measures the closeness between the 1-periphery node and the core members.
Participation Rate (Pr)
The Pr of a 1-periphery node is defined as the number of its connections to the core members over the total number of its connections. 0 < Pr ≤ 1. The Pr measures the level at which the 1-periphery members participate in the core. The Pr of closed single-core 1-periphery nodes is 1. The Pr of complete-open 1-periphery nodes are less than 0.5 and the Pr of limited-open 1-periphery nodes are greater than or equal to 0.5.
Peripheral Degree (Pd)
The Pd of a 1-periphery node is defined as the number of its connections to other 1-periphery nodes over the total number of peripheral nodes of the core . 0 ≤ Pd < 1. The Pd measures the degree to which the 1-periphery nodes are connected to each other. Pd = 0 means that the 1-periphery node is only connected to the core members.
Cores/peripheries are identified from the YPIN
Based on the criteria for the star structure, we identified 109 star cores with at least five 1-periphery nodes from the YPIN. Additional file 5 lists all star core proteins and their 1-peripheries and 2-peripheries.
Biological properties are different among cores and 1-peripheries
Comparing biological properties of k-plex core proteins and their 1-peripheries
Comparison of properties of k-plex core members with those of different types of 1-peripheries.
Number of domains
Complete open peripheries
Limited open peripheries
P value (core vs. complete open)
P value (core vs. limited open)
P value (core vs. closed)
P value (core vs. multiple)
Comparing biological properties of star core proteins and their 1-peripheries
Comparison of properties of star cores with those of their 1- peripheries and k-plex cores.
Number of domains
P value (star vs. 1-p.)
P value (star vs. k-plex)
Comparing biological properties of star core proteins and k-plex core proteins
We compared the biological properties of star cores and those of k-plex cores. As shown in Table 2, the average PCCs, and essentiality of k-plex core proteins are significantly different from those of star cores.
The k-plex cores can consist of party proteins, date proteins, or both
In order to examine the differences among k-plex core proteins, we analyzed how the k-plex core members related to the date/party concepts of Han et al. . Five microarray data sets [32–36] were used to determine the date and party classification of core proteins (see Methods for details). As a result (see Additional file 6), among the 706 proteins with PCCs in k-plex cores, 177 are party proteins and 529 are date proteins. Meanwhile, all star cores are date proteins except three. Moreover, lowering the threshold with significance level to 75% will not affect the conclusion that most k-plex core proteins are date proteins. This is a surprise, as the date core proteins are inside the functional modules (complex), rather than the external connectors . The party and date core proteins have similar degree (average 13.435 vs. 13.762). The Student's t-test has shown no significant difference on degree distribution. However, the clustering coefficients  of party core proteins (0.5824) are significantly higher than those of date core proteins (0.4259). We then classified the k-plex cores according to the party and date proteins inside them. A party core consists entirely of party proteins. A date core, on the other hand, consists entirely ofdate proteins. A mix core will include both party and date proteins. There are only 7 party cores, 37 mix cores and 66 date cores (see Additional file 7). This classification implied that the formation and evolution of protein complexes may involve different mechanisms.
The negative correlation between evolutionary rate and connectivity are much stronger among k-plex core members than among star core members
Expression dynamics are different among different kinds of links connecting core proteins and 1-periphery proteins
To get further insight to the expression difference between k-plex core proteins and their 1-periphery proteins, we compared the average PCCs of microarray expressions between two core proteins and between one core protein and one 1-periphery protein (see Methods for detailed calculation). For each k-plex core, the Additional file 8 listed the average PCCs for links between k-plex core members and for links between k-plex core members and their 1-peripheries. The overall average PCC for links between k-plex core members is 0.2532. And the overall average PCC for links between k-plex core members and their 1-peripheries is 0.1799. Two-tail T test on the average PCCs between two kinds of links shows significant difference (p-value = 1.74E-3).
Structural characteristics of 1-periphery proteins of k-plex cores imply two classes of 1-periphery proteins
Average of three structural measures of 1-peripheries
Furthermore, the Pd of all types of 1-periphery proteins are very small, 0.0724 for core-member 1-peripheries and less than 0.022 for other types of 1- peripheries. The low Pd of 1-peripheries indicates that 1-peripheries are generally not connected to each other. Thus, the peripheral members of the protein complexes usually may be assembled at different times and may involve distinct biological functions.
The Pr measures how the 1-periphery proteins connect to core members. All participation rates of closed single-core 1-periphery proteins are 1. The average participation rate of limited-open-single-core 1-periphery proteins is 0.5458. The average participation rates of complete-open, multiple-core, and core-member 1-periphery proteins are small, which indicates that the level that these three types of 1-peripheries associate with the cores is low. The high participation rates of closed and limited-open 1-periphery proteins indicate that they are more likely to join the protein complex (not the core). The low participation rates of multiple and complete-open 1-periphery proteins indicate that they are more likely to participate in different functionality or processes as they are connected to different complexes or individual proteins.
Therefore, we propose that closed and limited-open are "party periphery" proteins and multiple and complete-open are "connector periphery" proteins. As shown in Figure 2, most closed and limited-open 1-periphery proteins of the largest k-plex core are also part of the ribosome complex. Furthermore, we compared four properties, evolutionary rate, protein essentiality, number of domain, and average PCC, between the party periphery proteins and connector periphery proteins. The student's t-test showed that the party periphery proteins are significantly different from connector periphery proteins in all four properties, number of domains (p-value = 3.59E-4), essentiality (p-value = 6.40E-5), average PCCs (p-value = 1.74E-2) and evolutionary rate (p-value = 9.27E-3).
Both connectivity and topology are related to the property differences between k-plex core proteins and their 1-peripheral proteins
Correlation between connectivity and biological properties for different types of proteins
Number of domains
Complete open peripheries
Limited open peripheries
On the other hand, as shown in Table 4, there is no statistically significant correlation found in the following pair-wise comparisons: (1) between average PCC and connectivity for k-plex core proteins and "connector periphery" proteins; (2) between evolutionary rates and connectivity for "parity periphery" proteins; (3) between essentiality and the connectivity for all kinds of proteins rather than complete-open 1-periphery proteins; and (4) between number of domains and the connectivity for all kinds of proteins. Thus, in these cases, the topological types instead of connectivity may contribute to the difference in biological properties between nodes. In summary, both connectivity and topology are shown to be associated to the biological properties of proteins.
In this paper, we systematically explored the core/periphery structures in YPIN. We have identified 110 k-plex cores. Gene ontology based analysis showed that the 1-periphery proteins are closely related to the k-plex core proteins. However, low average coreness values of 1-periphery proteins indicated that peripheral proteins are structurally different from the k-plex core proteins. Furthermore, the properties of 1-peripheral proteins are significantly different from those of k-plex core proteins. Thus, it is meaningful to separate peripheral proteins from k-plex core proteins.
Based on their structural relationship with core members, we classified the non-core 1-periphery proteins into four types: closed-, limited-open, complete-open and multiple-core 1-periphery proteins. The closed and limited-open 1-periphery proteins, which have high participation rates, are structurally "party periphery" proteins. The complete-open and multiple-core 1-periphery proteins, which have low participation rates, are structurally "connector periphery" proteins. This classification may help understand different roles of 1-peripheiral proteins relate to the complex core. The "party periphery" 1-peripheral proteins are usually closely related to functionality of protein complex. On the other hand, the "connector periphery" 1-peripheral proteins are connectors that link the complex to other complexes or individual proteins.
Our results showed that the topological structures characteristics of proteins in PINs are reflected in their biological properties. For example, the closed and limited-open 1-periphery proteins have very similar topological structures and also have very similar biological properties. Furthermore, our results showed that, besides the connectivity, other structural characteristics are also related to biological properties. Thus, it is not enough to differentiate proteins based on connectivity only. Moreover, our studies showed that structure-properties relationship may be needed to take further analysis. For example, by further examining the relationship between the evolutionary rate and connectivity, we showed that there are differences between k-plex core proteins and star proteins.
The studies on the core/periphery structures in protein networks have also helped reveal expression dynamic difference in protein complexes [21, 22, 41]. The average PCC values of k-plex core members are significantly higher than those of their 1-peripheires. Furthermore, the average PCC values of links between k-plex core members are significantly higher than those of links between k-plex core members and their 1-periphery proteins. This dynamic difference implies the temporal "plug-and-play" components of protein complexes join the complexes after their formation.
We have compiled a yeast PIN (YPIN) by combining three curated yeast PINs: "Filtered Yeast Interactome" (FYI) , the Structure Interaction Network (SIN) , and the yeast core PIN downloaded from the DIP database (version ScereCR20070707) . After removal of all self-connecting links, the combined YPIN included 2,945 yeast proteins and 8,421 interactions. We applied our analysis to the single large component with 2,664 interconnected proteins (8,161 links) of this YPIN.
Algorithm for identifying k-plex cores
Borgatti and Everett  developed a genetic algorithm to separate small social networks into one core and its periphery. However, Boyd et al.  found that the Bett algorithm does not give the optimal results in most test cases. Rather, Boyd et al. found that the Kernighan-Lin (KL)  algorithm performs better in partitioning social networks into a core set and a periphery set. Here, we adapt the KL algorithm to identify all k-plex cores in the PINs.
The rationale behind this is that we would like to favor edges between core members and penalize disconnections between core members. This gain-based approach will result in a k-plex (k ≤ ⌊n/2⌋) with all core members having positive scores.
The choice of value of the k parameter in the k-plex definition will affect the cohesion of the sub-network. If k is too small, the sub-network will be too cohesive and exclude some loosely connected core members. On the other hand, if k is too large, the sub-network will be too loose and include some peripheral nodes into the core. The choice of k ≤ ⌊n/2⌋ in our k-plex core definition was selected to balance these interrelationships such that each core member must connect to at least half of the members in the core.
Our KL-like algorithm will start from each node triangle in the PIN. Each cycle, the algorithm will move nodes from periphery to core or from core to periphery to create a new core. Then, the next moving cycle will continue with the new core and its periphery. This procedure will continue until no new cores can be generated. The KL-like algorithm employs a greedy moving mechanism. Every time, the algorithm will move one node with the maximal gain among nodes in the core and its periphery. If a new core with a higher gain is obtained after a move, it will be recorded.
For every triangle of nodes in the network
Set triangle as the current best core set
Set the current best core as current core set
Set the current best core as previous best core
For every node in current core set and its 1-periphery but not in initial triangle
For every node in current core set and its 1-periphery but not in initial triangle
If it has not been moved
Calculate the gain of moving
Move the node with best gain
If the score of the current core is higher than the current best core
Store current core as the current best core
While the current best core has a greater score than the previous best core
Prune the results and remove the replicate cores.
Algorithm for identifying star structures
The procedure for locating star cores begins with finding all nodes that are not members of any k-plex core and have degree of at least five. Once these potential star nodes have been found, the periphery degrees of their 1-periphery nodes are examined. If the periphery degrees of all 1-peripheral nodes are greater than 0.16, the star node is kept. All star nodes that pass the examination are accepted as valid star cores. The reason that we choose 0.16 as the threshold for periphery degree is because it is possible to have stars with clustering coefficients greater than 0.1 if the maximum periphery degree threshold is beyond 0.16.
Determination of party and date proteins
Five microarray data sets [32–36] were downloaded from the Yeast Functional Genomics Database (YFGdb). All five data sets have at least 50 experiment data points, which should ensure the accuracy of the calculation of PCC. For each data set, the average PCC of each protein is obtained by averaging the PCC between the proteins and their neighbour proteins. Unfortunately, plotting the probability distribution of average PCC showed no clear bimodal distribution for all five data sets. Similar observations have been obtained by Ekman et al. . Instead of arbitrarily assigning a threshold, we modelled the distribution of average PCC as a normal distribution and calculated the mean and standard deviation of the distribution. We determined the threshold that separate "party" and "date" protein as the value that 90% of average PCC is below. Namely, the threshold for each microarray data set is the mean of average PCCs plus 1.282 times standard deviation. Addition file 6 lists the thresholds for all five microarray data sets. "Party" proteins are defined as proteins that have an average PCC from any of five microarray data sets higher than the threshold of that microarray data set; otherwise it is a "date" protein. Noted that we did not just classify hubs (with more than 5 links), but all core proteins.
Calculation of average PCCs of proteins
For each data set, the average PCC of each protein is obtained by average the PCC between the proteins and its neighbour proteins. Then, the average PCC of a protein is calculated by averaging the five average values from the five data sets.
Calculation of average PCC of links
For each data set, the PCC between each pair of connected proteins is calculated. The PCC of a link is the average PCC from the five data sets.
Feng Luo and Bo Li are supported by NSF EPSCoR grant EPS-0447660. Richard Scheuermann is supported by the National Institutes of Health contracts N01-AI40076 and N01-AI40041. XWF is partially supported by NSF Award BCS-0717688.
This article has been published as part of BMC Bioinformatics Volume 10 Supplement 4, 2009: Proceedings of the IEEE International Conference on Bioinformatics and Biomedicine (BIBM) 2008. The full contents of the supplement are available online at http://www.biomedcentral.com/1471-2105/10?issue=S4.
- Barabasi A-L, Oltvai ZN: Network biology: understanding the cell's functional organization. Nature Reviews Genetics 2004, 5(2):101–113. 10.1038/nrg1272View ArticlePubMedGoogle Scholar
- Barabási A-L, Albert R: Emergence of Scaling in Random Networks. Science 1999, 286(5439):509–512. 10.1126/science.286.5439.509View ArticlePubMedGoogle Scholar
- Hartwell LH, Hopfield JJ, Leibler S, Murray AW: From molecular to modular cell biology. Nature 1999, 402(6761):C47-C52. 10.1038/35011540View ArticlePubMedGoogle Scholar
- Ravasz E, Somera AL, Mongru DA, Oltvai ZN, Barabasi AL: Hierarchical Organization of Modularity in Metabolic Networks. Science 2002, 297(5586):1551–1555. 10.1126/science.1073374View ArticlePubMedGoogle Scholar
- Watts DJ, Strogatz SH: Collective dynamics of 'small-world' networks. Nature 1998, 393(6684):440–442. 10.1038/30918View ArticlePubMedGoogle Scholar
- Jeong H, Mason SP, Barabasi AL, Oltvai ZN: Lethality and centrality in protein networks. Nature 2001, 411(6833):41–42. 10.1038/35075138View ArticlePubMedGoogle Scholar
- Ito T, Chiba T, Ozawa R, Yoshida M, Hattori M, Sakaki Y: A comprehensive two-hybrid analysis to explore the yeast protein interactome. Proceedings of the National Academy of Sciences of the United States of America 2001, 98(8):4569–4574. 10.1073/pnas.061034498PubMed CentralView ArticlePubMedGoogle Scholar
- Uetz P, Giot L, Cagney G, Mansfield TA, Judson RS, Knight JR, Lockshon D, Narayan V, Srinivasan M, Pochart P, et al.: A comprehensive analysis of protein-protein interactions in Saccharomyces cerevisiae. Nature 2000, 403(6770):623–627. 10.1038/35001009View ArticlePubMedGoogle Scholar
- Guimera R, Sales-Pardo M, Amaral LAN: Classes of complex networks defined by role-to-role connectivity profiles. Nature Physics 2007, 3(1):63–69. 10.1038/nphys489PubMed CentralView ArticlePubMedGoogle Scholar
- Newman ME, Girvan M: Finding and evaluating community structure in networks. Physical review 2004, 69(2 Pt 2):026113.PubMedGoogle Scholar
- Fortunato S, Barthelemy M: Resolution limit in community detection. Proceedings of the National Academy of Sciences of the United States of America 2007, 104(1):36–41. 10.1073/pnas.0605965104PubMed CentralView ArticlePubMedGoogle Scholar
- Wang Z, Zhang J: In Search of the Biological Significance of Modular Structures in Protein Networks. PLoS Comput Biol 2007, 3(6):e107. 10.1371/journal.pcbi.0030107PubMed CentralView ArticlePubMedGoogle Scholar
- Mewes HW, Frishman D, Mayer KF, Munsterkotter M, Noubibou O, Pagel P, Rattei T, Oesterheld M, Ruepp A, Stumpflen V: MIPS: analysis and annotation of proteins from whole genomes in 2005. Nucleic acids research 2006, (34 Database):D169–172. 10.1093/nar/gkj148Google Scholar
- Altaf-Ul-Amin M, Shinbo Y, Mihara K, Kurokawa K, Kanaya S: Development and implementation of an algorithm for detection of protein complexes in large interaction networks. BMC bioinformatics 2006, 7: 207. 10.1186/1471-2105-7-207PubMed CentralView ArticlePubMedGoogle Scholar
- Bader GD, Hogue CW: An automated method for finding molecular complexes in large protein interaction networks. BMC bioinformatics 2003, 4: 2. 10.1186/1471-2105-4-2PubMed CentralView ArticlePubMedGoogle Scholar
- Hu H, Yan X, Huang Y, Han J, Zhou XJ: Mining coherent dense subgraphs across massive biological networks for functional discovery. Bioinformatics (Oxford, England) 2005, 21(Suppl 1):i213–221. 10.1093/bioinformatics/bti1049View ArticleGoogle Scholar
- Spirin V, Mirny LA: Protein complexes and functional modules in molecular networks. Proceedings of the National Academy of Sciences of the United States of America 2003, 100(21):12123–12128. 10.1073/pnas.2032324100PubMed CentralView ArticlePubMedGoogle Scholar
- Gavin AC, Aloy P, Grandi P, Krause R, Boesche M, Marzioch M, Rau C, Jensen LJ, Bastuck S, Dumpelfeld B, et al.: Proteome survey reveals modularity of the yeast cell machinery. Nature 2006, 440(7084):631–636. 10.1038/nature04532View ArticlePubMedGoogle Scholar
- Laumann EO, Pappi FU: Network of Collective Action: A Perspective on Community Influence Systems. Academics Press, New York; 1976.Google Scholar
- Borgatii SP, Everett MG: Models of core/periphery structures. Social Networks 1999, 21: 375–395. 10.1016/S0378-8733(99)00019-2View ArticleGoogle Scholar
- de Lichtenberg U, Jensen LJ, Brunak S, Bork P: Dynamic complex formation during the yeast cell cycle. Science 2005, 307(5710):724–727. 10.1126/science.1105103View ArticlePubMedGoogle Scholar
- Han JD, Bertin N, Hao T, Goldberg DS, Berriz GF, Zhang LV, Dupuy D, Walhout AJ, Cusick ME, Roth FP, et al.: Evidence for dynamically organized modularity in the yeast protein-protein interaction network. Nature 2004, 430(6995):88–93. 10.1038/nature02555View ArticlePubMedGoogle Scholar
- Fraser HB: Modularity and evolutionary constraint on proteins. Nature genetics 2005, 37(4):351–352. 10.1038/ng1530View ArticlePubMedGoogle Scholar
- Fraser HB, Hirsh AE, Steinmetz LM, Scharfe C, Feldman MW: Evolutionary rate in the protein interaction network. Science 2002, 296(5568):750–752. 10.1126/science.1068696View ArticlePubMedGoogle Scholar
- Everett MG, Borgatii SP: Peripheries of cohesive subsets. Social Networks 1999, 21: 397–407. 10.1016/S0378-8733(99)00020-9View ArticleGoogle Scholar
- Luce RD, Perry A: A method of marix analysis of group structure. Psychometrika 1949, 14: 94–116. 10.1007/BF02289146View ArticleGoogle Scholar
- Seidman SB, Foster BL: A Grpah-theoretic generalization of the clique concept. Journal of Mathematical sociology 1978, 6: 139–154.View ArticleGoogle Scholar
- Bollobas B: The evolution of sparse graphs. Graph theory and combinatorics 1984, 35–57.Google Scholar
- Luce RD: Conectivity and generalized cliques in a sociometric group structure. Psychometrika 1950, 15: 159–190. 10.1007/BF02289199View ArticleGoogle Scholar
- Cherry JM, Adler C, Ball C, Chervitz SA, Dwight SS, Hester ET, Jia Y, Juvik G, Roe T, Schroeder M, et al.: SGD: Saccharomyces Genome Database. Nucleic acids research 1998, 26(1):73–79. 10.1093/nar/26.1.73PubMed CentralView ArticlePubMedGoogle Scholar
- Gasch AP, Huang M, Metzner S, Botstein D, Elledge SJ, Brown PO: Genomic expression responses to DNA-damaging agents and the regulatory role of the yeast ATR homolog Mec1p. Molecular biology of the cell 2001, 12(10):2987–3003.PubMed CentralView ArticlePubMedGoogle Scholar
- Gasch AP, Spellman PT, Kao CM, Carmel-Harel O, Eisen MB, Storz G, Botstein D, Brown PO: Genomic expression programs in the response of yeast cells to environmental changes. Molecular biology of the cell 2000, 11(12):4241–4257.PubMed CentralView ArticlePubMedGoogle Scholar
- Huisinga KL, Pugh BF: A TATA binding protein regulatory network that governs transcription complex assembly. Genome biology 2007, 8(4):R46. 10.1186/gb-2007-8-4-r46PubMed CentralView ArticlePubMedGoogle Scholar
- O'Rourke SM, Herskowitz I: Unique and redundant roles for HOG MAPK pathway components as revealed by whole-genome expression analysis. Molecular biology of the cell 2004, 15(2):532–542. 10.1091/mbc.E03-07-0521PubMed CentralView ArticlePubMedGoogle Scholar
- Spellman PT, Sherlock G, Zhang MQ, Iyer VR, Anders K, Eisen MB, Brown PO, Botstein D, Futcher B: Comprehensive identification of cell cycle-regulated genes of the yeast Saccharomyces cerevisiae by microarray hybridization. Molecular biology of the cell 1998, 9(12):3273–3297.PubMed CentralView ArticlePubMedGoogle Scholar
- Wall DP, Hirsh AE, Fraser HB, Kumm J, Giaever G, Eisen MB, Feldman MW: Functional genomic analysis of the rates of protein evolution. Proceedings of the National Academy of Sciences of the United States of America 2005, 102(15):5483–5488. 10.1073/pnas.0501761102PubMed CentralView ArticlePubMedGoogle Scholar
- Giaever G, et al.: Functional profiling of the Saccharomyces cerevisiae genome. Nature 2002, 418(6896):387–391. 10.1038/nature00935View ArticlePubMedGoogle Scholar
- Saeed R, Deane CM: Protein protein interactions, evolutionary rate, abundance and age. BMC Bioinformatics 2006, 7: 128. 10.1186/1471-2105-7-128PubMed CentralView ArticlePubMedGoogle Scholar
- Bloom JD, Adami C: Apparent dependence of protein evolutionary rate on number of interactions is linked to biases in protein-protein interactions data sets. BMC Evolutionary Biology 2003, 3: 21. 10.1186/1471-2148-3-21PubMed CentralView ArticlePubMedGoogle Scholar
- Komurov K, White M: Revealing static and dynamic modular architecture of the eukaryotic protein interaction network. Molecular systems biology 2007, 3: 110. 10.1038/msb4100149PubMed CentralView ArticlePubMedGoogle Scholar
- Kim PM, Lu LJ, Xia Y, Gerstein MB: Relating three-dimensional structures to protein networks provides evolutionary insights. Science 2006, 314(5807):1938–1941. 10.1126/science.1136174View ArticlePubMedGoogle Scholar
- Deane CM, Salwinski L, Xenarios I, Eisenberg D: Protein Interactions: Two Methods for Assessment of the Reliability of High Throughput Observations. Mol Cell Proteomics 2002, 349–356.Google Scholar
- Boyd JP, Fitzgerald WJ, Beck RJ: Computing core/periphery structures and permutation tests for social relations data. Social Networks 2006, 28: 166–178. 10.1016/j.socnet.2005.06.003View ArticleGoogle Scholar
- Kernighan BW, Lin S: An efficient heuristic procedure for partitioning graphs. Bell System Technical Journal 1970, 49: 221–226.View ArticleGoogle Scholar
- Ekman D, Light S, Bjorklund AK, Elofsson A: What properties characterize the hub proteins of the protein-protein interaction network of Saccharomyces cerevisiae? Genome Biology 2006, 7(6):R45. 10.1186/gb-2006-7-6-r45PubMed CentralView ArticlePubMedGoogle Scholar
- Enright AJ, Ouzounis CA: BioLayout–an automatic graph layout algorithm for similarity visualization. Bioinformatics (Oxford, England) 2001, 17(9):853–854. 10.1093/bioinformatics/17.9.853View ArticleGoogle Scholar
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 (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.