- Research article
- Open Access
Stable evolutionary signal in a Yeast protein interaction network
BMC Evolutionary Biologyvolume 6, Article number: 8 (2006)
The recently emerged protein interaction network paradigm can provide novel and important insights into the innerworkings of a cell. Yet, the heavy burden of both false positive and false negative protein-protein interaction data casts doubt on the broader usefulness of these interaction sets. Approaches focusing on one-protein-at-a-time have been powerfully employed to demonstrate the high degree of conservation of proteins participating in numerous interactions; here, we expand his 'node' focused paradigm to investigate the relative persistence of 'link' based evolutionary signals in a protein interaction network of S. cerevisiae and point out the value of this relatively untapped source of information.
The trend for highly connected proteins to be preferably conserved in evolution is stable, even in the context of tremendous noise in the underlying protein interactions as well as in the assignment of orthology among five higher eukaryotes. We find that local clustering around interactions correlates with preferred evolutionary conservation of the participating proteins; furthermore the correlation between high local clustering and evolutionary conservation is accompanied by a stable elevated degree of coexpression of the interacting proteins. We use this conserved interaction data, combined with P. falciparum /Yeast orthologs, as proof-of-principle that high-order network topology can be used comparatively to deduce local network structure in non-model organisms.
High local clustering is a criterion for the reliability of an interaction and coincides with preferred evolutionary conservation and significant coexpression. These strong and stable correlations indicate that evolutionary units go beyond a single protein to include the interactions among them. In particular, the stability of these signals in the face of extreme noise suggests that empirical protein interaction data can be integrated with orthologous clustering around these protein interactions to reliably infer local network structures in non-model organisms.
An ambitious goal of contemporary proteome research is the elucidation of the structure, interactions and functions of the proteins that constitute cells and organisms. During the last few years, large-scale efforts have unraveled the complex web of protein interactions in simple organisms such as H. pylori , E. coli  and S. cerevisiae [3–7]. Most recently, attention has focused on the first protein interaction maps of complex multicellular organisms such as C. elegans  and D. melanogaster . Although these organisms vary extensively in their complexity, corroborative evidence points to a series of simple organizing principles that characterize all complex protein interaction networks . The most dramatic of these is their scale-free nature [11, 12], highlighting a small number of highly connected proteins which secure the integrity and connectivity among modules [13, 14] that are discernible, yet topologically overlapping, clusters of densely interconnected protein groups sharing well-defined functions [10, 15–18]. A crucial biological corollary of this ubiquitous network organization is the observation that hubs exhibit an elevated propensity to be simultaneously conserved in evolution and are essential for survival [13, 19, 20]. This role of highly connected proteins is further indicated by a considerable degree of sequence conservation [21–25]. Similarly, cohesively bound modules have been conserved as a whole, suggesting the presence of evolutionary relevant building blocks [26–28]. This hypothesis is further supported by the observation that proteins belonging to a certain module tend to be coexpressed  and coregulated . These particular results are utilized for the comparison of protein pathways of various organisms , modeling of interactomes [32, 33] and prediction of protein functions .
These insights have fundamental implications for our understanding of biological processes and potential applications; however the severe error-proneness of methods for the determination of protein interactions casts doubt on the integrity of such datasets. For example, an estimate of the accuracy of protein interactions in S. cerevisae uncovered a startling false negative rate of 90%, and a 50% false positive error rate .
Despite incoherences in the determination of protein interactions and orthologs, we observe that extensive information remains in the topology of a protein interaction network. In particular, even tremendous experimental noise does not bury the strong evolutionary signal that highly connected nodes in an interaction web of Yeast proteins are preferably conserved in higher eukaryotes. Accounting for interactions between pairs of Yeast proteins, we find that the reliability of an interaction as indicated by a high degree of local clustering around interactions is accompanied by an elevated propensity for the corresponding proteins to be evolutionary conserved. In addition, we observe that such interactions are preferably coexpressed in both the reference and a target organism, suggesting that conservation occurs not only on the level of individual proteins but also on the level of their interactions. The observation that such link-based evolutionary signals prevail in the topology of an otherwise extremely noisy protein interaction network indicates a novel way to uncover protein interactions in any organism for which orthologs can be identified from sequence data.
As a basis of our considerations we utilized a protein-protein interaction network of S. cerevisae from the DIP database , providing 3, 833 proteins embedded in 11, 942 interactions. We labeled pairs of proteins as orthologous to each other as of the InParanoid database  that relates proteins of S. cerevisiae to complete protein sets of various higher eukaryotes, allowing us to utilize 1, 928 Yeast proteins with putative orthologs in H. sapiens, 2,073 in A. thaliana, 1, 885 in C. elegans, 1, 885 in M. musculus and 1,631 in D. melanogaster.
Evolutionary retention of single proteins
Utilizing these data sets we recently uncovered a correlation between a Yeast proteins level of interaction and its propensity to be evolutionary conserved . Pooling all proteins into groups according to their connectivity k we determine the respective fraction of orthologs in each group. As a null-hypothesis we assume a random distribution of orthologs that is quantified by the fraction of proteins with an ortholog in a target eukaryote and the total number of proteins present in the underlying Yeast protein interaction network. The degree dependent orthologous excess retention, ER k , defined as the ratio of ortholog fractions in k dependent groups of proteins and fractions of randomly distributed orthologous proteins reflects the dependence of evolutionary protein conservations as function of the proteins connectivity. Logarithmically binning the k-dependent values of ER k the averages in each bin show a clear and systematic trend toward preferred conservation of proteins that interact on a high level (Fig. 1a). Significant Pearson's and Spearman's rank coefficients support our qualitative observations [see Additional file 1].
Evolutionary retention of interacting pairs of proteins
While we find that the conservation of single proteins is a function of connectedness we wonder if topology also contains such evolutionary signals on the level of interactions. Because proteins which are placed in cohesive areas (i.e. modules) tend to be evolutionary conserved we wonder if their interactions are conserved too. We utilize a link-based clustering coefficient that reflects the degree of clustering of an interaction's immediate network neighborhood, a topological measure that allows for correlations between local clustering and the actual reliability of observed interactions . Similar to the single protein case, we grouped all interactions according to their hypergeometric clustering coefficient C vw and determined the respective fraction of interacting pairs that are fully conserved as putative orthologs in each bin. In the absence of a correlation between evolutionary conservation and an interactions placement in the network the ratio of the C vw -dependent and random fractions of orthologous protein pairs – defined as the interaction based excess retention (see Materials and Methods) – would be unity. Logarithmically binning all interactions according to their local degree of clustering C vw and determining the average excess retention in each bin we identify a significant and systematic trend of proteins engaged in highly clustered interactions to be preferably evolutionary conserved [Fig. 1b, see Additional file 1]. These link-based observations are not only consistent with previous node-based results but also allow to suggest that standard single-node measurements of evolutionary conservation can be extended to their neighboring links. This evolutionary corollary indicates that not only single proteins are a target of evolution but also the interactions between conserved proteins.
To demonstrate this gain of evolutionary information, we simulated the impact of extremely high false negatives rates of protein interactions by removing up to 70% of experimentally determined links between randomly selected protein pairs. Additionally, to address the effects of false positives, we randomly distributed up to 70% more interactions than were previously identified in the original Yeast network.
Because there are no significant differences in the distributions of organism-specific excess retention in Fig. 1, we examine orthologs of C. elegans as a representative comparative set for these analyzes. After generating 1, 000 different realizations to each case of incomplete false data, we determine the excess retention ER of proteins and their interactions that have orthologs in C. elegans according to their degree k and hypergeometric clustering coefficient C vw . Determining the average excess retention in bins of increasing logarithmic size we find that the relationship between excess retention, level of interaction and local clustering is widely unaltered (single proteins: Figs. 2a,b; protein interactions: Figs. 2c,d). Addressing the statistical significance of these results we determined correlation coefficients (Pearson's r and Spearman's rank ρ) and performed Kolmogorov-Smirnov tests, recovering significant similarities between the original and perturbed distributions [see Additional file 1].
Moreover, to represent missed orthologs, we randomly eliminated up to 70% from the set of Yeast proteins that have an ortholog in C. elegans. In turn, we randomly labeled up to 70% more proteins as orthologs in C. elegans than were previously present in the initial set. Sampling 1, 000 different realizations each, we calculated the excess retention according to the proteins degree k and local clustering around each interaction C vw . Logarithmically binning the results thus obtained we averaged the excess retention of orthologous proteins in each bin, allowing us to find that the introduction of noise on the level of orthologs determination does not alter our initial observations (single proteins: insets Figs. 2a,b; protein interactions: insets Figs. 2c,d). Significantly similar correlation coefficients and Kolmogorov-Smirnov scores [see Additional file 1] support our conclusions.
Clustering, coexpression and evolutionary conservation
The observation that highly clustered links between evolutionarily conserved proteins are reliable and stable toward severe perturbation enhances our expectation that an elevated degree of coexpression of interacting proteins will retain this relationship as well. In particular, a strong coexpression signal of the orthologs of proteins that embrace the interactions in question would strongly indicate the actual presence of the interaction in a reference and target organism. As a test case, we extend our investigations to the malaria parasite Plasmodium falciparum, a single celled organism that has 895 putative orthologous proteins with Yeast. Elucidating those Yeast interactions between yeast proteins conserved in P. falciparum, we find a web of 3, 071 interactions among 659 proteins in P. falciparum. To evaluate the quality of these inferred interactions we utilized a comprehensive set of P. falciparum specific coexpression data  to calculate Pearson's correlation coefficients r P for the inferred protein interactions. In the same way, we utilized an extensive set of Yeast coexpression data  to investigate the coexpression tendency of those interacting Yeast proteins that served as the template for the inferred interactions in Plasmodium. In both cases, we use the Yeast specific C vw values as an approximate measure of an interaction's reliability. Focusing on interactions that score above increasing thresholds of C vw , we observe a strong shift toward coexpression of the considered protein pairs (Fig. 3a,b). The difference between the individual coexpression patterns is further indicated by significant Students t-test scores when the C vw dependent distributions are compared to a background distribution of coexpression coefficients of all protein pairs in the considered organism [see Additional file 1]. The significant shifts toward elevated levels of coexpression identify a pronounced correlation between the local cohesiveness of an interaction and the tendency that the involved proteins are coexpressed. The determination of mean coexpression coefficients r P of interactions that have been logarithmically grouped according to their C vw allows us to find a statistically significant trend toward elevated levels of coexpression of conserved yeast interactions that are placed in highly clustered neighborhood (inset, Fig. 3c). Although we adopt measurements of the local cohesiveness around links C vw from Yeast, we find that the corresponding interactions in Plasmodium exhibit a similar trend (Fig. 3c). Remarkably, the latter distribution exhibits better and statistically more significant correlation coefficients than its template in Yeast [see Additional file 1].
In the same way we investigated the stability of the interactions propensity to be evolutionary conserved, we checked for the robustness of the obtained correlation between local clustering and coexpression. Mimicking the presence of false positive/negative links we randomly eliminated/added up to 70 % of interactions in the Yeast interaction network. Recalculating the hypergeometric clustering coefficient for each of 1,000 runs, we grouped all interacting pairs of Yeast proteins with an ortholog in P. falciparum according to C vw in bins of logarithmically increasing size. Averaging over the respective coexpression correlation coefficient r P of all Yeast interactions in each bin, we observe that the initial ascending trend prevails [see Fig. 1ab of Additional file 1]. Assuming that all interactions between proteins that have an ortholog in Plasmodium were conserved we repeated this procedure by superimposing and averaging over the respective coexpression correlation values of Plasmodium. Similar to the Yeast specific case, we observe the same qualitative trends [see Fig. 2ab of Additional file 1]. Significant correlation coefficients and Kolmogorov-Smirnov scores support our observations of our findings. In the same way, we simulated the presence of false positive /negative orthologs by eliminating/adding up to 70% of orthologs in P. falciparum. Averaging over 1,000 runs each, we determined the average coexpression coefficient in each bin utilizing Plasmodium and Yeast specific coexpression data. In each case, we find that the original trend of high local clustering around interactions coincides with an increased propensity to be coexpressed strongly prevails [see Figs. 1, 2bc of Additional file 1], observations that are supported by significant correlation coefficients and Kolmogorov-Smirnov scores [see Additional file 1].
Discussion & conclusion
Extending a previous study indicating that highly interacting proteins are predominantly conserved in evolution we generalize the concept that evolutionary signals are carried by the topology of the underlying protein interaction network. In particular, a protein's propensity to be conserved while interacting with a high number of partners – a node-based evolutionary signal – has a link based counterpart, as indicated by the propensity of interacting proteins to be evolutionary conserved with increasing local clustering around the interaction in question. Although the obtained correlations are significant, the alarmingly high error rates in the determination of protein interactions cast doubt on the obtained results.
By focusing on perturbation events on node and interaction levels, we observe that extreme error rates of both protein interactions and orthologs do not ablate the evolutionary signal carried by the network structure. The introduction of noise at the node, by simulation of inconsistent determination of orthologs, does not override the preference of highly connected nodes to be evolutionary conserved; as theoretically predicted, random perturbations will rarely affect a hub in a scale-free network . The low probability that a hub is hit by a random perturbation event also explains that interacting proteins that are placed in a highly clustered environment retain their evolutionary signal. Indeed, the definition of the hypergeometric clustering coefficient assures a high score for interacting proteins that share a lot of their interaction partners.
On an interaction level, we observe that the massive insertion/deletion of links does not obliterate the local structure of networks as indicated by the stable preference of highly connected proteins and protein pairs that are embedded in a well clustered neighborhood to be evolutionary conserved. In particular, we conclude that insertion/deletion of random links on average impact sparsely connected parts of the networks much more than densely connected ones; indeed, loss of information in highly clustered neighborhoods and highly connected hubs would require massive, targeted deletion/insertion of links to obliterate their local structure. Therefore, the observation that links which are placed in a highly clustered neighborhood are highly reliable  is nested in our observation that highly clustered neighborhoods compensate severe random perturbations much better than sparsely connected ones.
While our results allow us to conclude that degree alone is a robust indicator for a proteins propensity to be evolutionary conserved, the inherent topological robustness of locally clustered links emphasizes the emergent role of cohesive areas  as mediators of evolutionary information. In the simplest case, we confirmed that not only single proteins are a potential target of evolution but interaction among them can be potentially conserved as well. As a strong indicator that an interaction indeed has been conserved, the correlation between high local clustering and evolutionary conservation is accompanied by a stable elevated degree of coexpression of the interacting proteins in both a model and target organism. Superimposing the extreme error rates simulating the incoherent determination of orthologs and interactions as well we see that trends in both the model and target organism prevail, strongly indicating that evolution also happens on the level of interactions and putative bundles of interactions.
Although we utilized very noisy and inconsistent data of protein interactions and putative orthologs, we see that high connectivity and high clustering on average harbor significantly more evolutionary relevant information that sparsely connected and clustered areas. The coincidence of (i) high local clustering around highly reliable interactions of proteins, (ii) their propensity to be evolutionary conserved, (iii) their tendency to be coexpressed even in the face of tremendous experimental noise sketches a hypothetical framework to infer an evolutionary core of single protein-protein interactions by elucidating interacting proteins of a reference organism that have orthologs in the targeted organism. The quality of an interaction is assessed by calculating the corresponding hypergeometric clustering coefficient. Choosing the highest scoring – thus most reliable – ortholog interaction allows the selection of a core interaction network in the targeted organism. Unlike our case, where evolutionary relationships between proteins were approximated by similarity searches, the quality of predicted interactions will be enhanced by utilizing more sophisticated methods (such as tree-base methods) which allow a more reliable assignment of orthology. Finally, the cross-validation with high resolution coexpression data can refine specific protein-protein interaction subnetworks, allowing for checks of the actual presence of a proposed interaction. Ultimately, such a framework would allow a first insight into evolutionary conserved parts in interactomes of organism for which no interaction data currently exists.
As a source of protein interactions we chose the DIP database  which provides a set of manually curated protein-protein interactions in the organism S. cerevisiae. The current version contains 3, 833 proteins involved in 11, 942 interactions derived from combined, non-overlapping data which are mostly obtained from the high-throughput application of the two-hybrid method.
Assignment of orthology
Orthologs are genes in different species that originate from a single gene in the last common ancestor of these species. Such genes often have retained identical biological roles in present day organisms, indicated by a high degree of sequence homology. Unfortunately, orthology analysis between organisms is often difficult and error prone because of large numbers of paralogs within protein families. As a source of reliable and robust information about orthologous relationships between proteins in different species we utilized the InParanoid database [37, 42] which provides putative orthologous sequence information for S. cerevisiae and numerous other organisms. The algorithm for assigning orthologous relationships is based on pairwise similarity scores which are by default calculated with the BLASTP program. Best pairwise hits between the proteomes of two species are seeds – labeled as the main ortholog groups – of orthologous protein sequence clusters. In a further step, other sequences are added to this group if they are closely homologous to one of the main orthologs, members of orthologous groups which are called in-paralogs. In a final quality checking step, confidence values for each ortholog and in-paralog is determined allowing the detection of putative orthologous relationships that has been only reliably possible by multiple alignments and phylogenetic trees previously . In our study, we considered the main ortholog pairs of each orthologous group as sequences that are putatively orthologous to each other allowing us to obtain 1, 928 Yeast proteins with orthologs in H. sapiens, 2,073 in A. thaliana, 1, 885 in C. elegans, 1, 885 in M. musculus, 1,631 in D. melanogaster and 895 in P. falciparum.
Hypergeometric clustering coefficient
Recently, a network topology based approach uncovered a remarkable correlation between enhanced quality of protein interactions and the degree of clustering of their immediate network neighborhood . Considering a protein-protein interaction network with N nodes, we define the hypergeometric clustering coefficient as
where N(x) represents the neighborhood of a vertex x. Given fixed neighborhood sizes N(v) and N(w) of proteins v and w, the hypergeometric clustering coefficient increases with elevated overlap between the protein's neighborhoods. Provided that the neighborhoods are independent, the summation can be interpreted as a p value, reflecting the probability of obtaining a number of mutual neighbors between proteins v and w at or above the observed number by chance.
Orthologous excess retention
According to their hypergeometric clustering coefficient C vw of the interactions they are involved in, we grouped all interactions in groups of same C vw that have been rounded to integers. For each group of proteins, the fraction of interacting pairs of proteins that both have an ortholog in an other organism is defined as . In the absence of a correlation between evolutionary conservation of interacting protein pairs and their position in the network, has the general C vw -independent value e o = n o /N, where n o is the total number of interactions between Yeast proteins that have an ortholog, and N is the total number of Yeast protein interactions in the underlying network. Thus, we define the clustering-dependent excess retention of such proteins as which has the C vw -independent value for a random distribution of orthologous proteins . Basically, we applied the same framework for single proteins, by grouping them according to their degree k. For each group of N k proteins, the fraction of proteins that also have an ortholog is defined as ek,o= nk,o/N k . Analogously, the node based excess retention ER k is defined as ER k = ek,o/E k , where E k is the ratio of all proteins with an ortholog in the whole network.
To evaluate the quality of these inferred interactions we utilized a comprehensive set of Plasmodium specific  and Yeast specific  coexpression data. In each dataset, we utilized the expression profiles to determine the respective Pearson's correlation coefficient r P for each interacting pair of proteins.
To guarantee balanced sampling of our distributions we generally use logarithmic binning of the respective x-axis, a procedure for curve estimation that corrects for the skewed nature of the scale-free distribution.
On a logarithmic scale, we define the bin size , where N corresponds to the selected number of bins. Values a and b refer to the minimal and maximal value of data points on the x-axis, b = max i (x i ) and a = min i (x i ). Thus, n i = /Δ, n i ∈ [0, N - 1] reflects the number of the bin we assign a data point with a x i coordinate. Representing the n i th bin on the x-axis, we place at the end of each bin using .
The advantage of logarithmic binning is an elevated degree of noise reduction which is dependent on the bin size [41, 43]. Although this procedure causes a loss of accuracy, we still uncover the buried trends to a satisfying extent applying our statistical methods on the binned data.
Rain JC, Selig L, DeReuse H, Battaglia V, Reverdy C, Simon S, Lenzen G, Petel F, Wojcik J, Schächter V, Chemama Y, Labigne A, Legrain P: The protein-protein interaction map of Helicobacter pylori. Nature. 2001, 409: 211-215. 10.1038/35051615.
Butland G, Peregrin-Alvarez JM, Li J, Yang W, Yang X, Canadien V, Starostine A, Richards D, Beattie B, Krogan N, Davey M, Parkinson J, Greenblatt J, Emili A: Interaction network conatining conserved and essential protein complexes in Escherichia coli. Nature. 2005, 433: 531-537. 10.1038/nature03239.
Ito T, Tashiro K, Muta S, Ozawa R, Chiba T, Nishizawa M, Yamamoto K, Kuhara S, Sakaki Y: Towards a protein-protein interaction map of the budding yeast: A comprehensive system to examine two-hybrid interactions in all possible combinations between the yeast proteins. Proc Nat Acad Sci USA. 2000, 97 (3): 1143-1147. 10.1073/pnas.97.3.1143.
Schwikowski B, Uetz P, Fields S: A network of protein-protein interactions in yeast. Nature Biotechn. 2000, 18: 1257-1261. 10.1038/82360.
Uetz P, Giot L, Cagney G, Mansfield T, Judson R, Knight J, Lockshorn D, Narayan V, Srinivasan M, Pochart P, Qureshi-Emili A, Li Y, Godwin B, Conover D, Kalbfleisch T, Vijayadamodar G, Yang M, Johnston M, Fields S, Rothberg J: A comprehensive analysis of protein-protein interactions of Saccharomyces cerevisiae. Nature. 2000, 403: 623-627. 10.1038/35001009.
Gavin A, Bösche M, Krause R, Grandi P, Marzioch M, Bauer A, Schultz J, Rick J, Michon AM, Cruciat CM, Remor M, Böfert C, Schelder M, Brajenovic M, Ruffner H, Merino A, Klein K, Hudak M, Dickson D, Rudi T, Gnau V, Bauch A, Bastuck S, Huhse B, Leutwein C, Heurtier MA, Copley R, Edelmann A, Querfurth E, Rybin V, Drewes G, Raida M, Bouwmeester T, Bork P, Seraphin B, Kuster B, Neubauer G, Superti-Furga G: Functional organization of the yeast proteome by systematic analysis of protein complexes. Nature. 2002, 415: 141-147. 10.1038/415141a.
Ho Y, Gruhler A, Heilbut A, Bader G, Moore L, Adams SL, Millar A, Taylor P, Bennett K, Boutillier K, coauthors: Systematic identification of protein complexes in Saccharomyces cerevisiae by mass spectrometry. Nature. 2002, 415: 180-183. 10.1038/415180a.
Walhout A, Sordella R, Lu X, Hartley J, Temple G, Brasch M, Thierry-Mieg N, Vidal M: Protein interaction mapping in C. elegans using proteins involved in vulval development. Science. 2000, 287: 116-122. 10.1126/science.287.5450.116.
Giot L, Bader J, Brouwer C, Chaudhuri A, Kuang B, Li Y, Hao Y, Ooi C, Godwin B, Vitols E, Vijayadamodar G, Pochart P, Machineni H, Welsh M, Kong Y, Zerhusen B, Malcolm R, Varrone Z, Collis A, Minto M, Burgess S, McDaniel L, Stimpson E, Spriggs F, Williams J, Neurath K, Ioime N, Agee M, Voss E, Furtak K, Renzulli R, Aanensen N, Carrolla S, Bickelhaupt E, Lazovatsky Y, DaSilva A, Zhong J, Stanyon C, Finley R, White K, Braverman M, Jarvie T, Gold S, Leach M, Knight J, Shimkets R, McKenna M, Chant J, Rothberg J: A Protein Interaction Map of Drosophila melanogaster. Science. 2004, 302: 1727-1736. 10.1126/science.1090289.
Barabaśi A, Oltvai Z: Network Biology: Understanding the Cell's Functional Organization. Nature Rev Gen. 2004, 101-113. 10.1038/nrg1272. 5
Barabási A, Albert R: Emergence of Scaling in Random Networks. Science. 1999, 286: 509-512. 10.1126/science.286.5439.509.
Albert R, Barabási AL: Statistical mechanics of complex networks. Rev Mod Phys. 2002, 74: 47-10.1103/RevModPhys.74.47.
Jeong H, Mason S, Barabási AL, Oltvai Z: Lethality and centrality in protein networks. Nature. 2001, 411: 41-42. 10.1038/35075138.
Han J, Bertin N, Hao T, Goldberg DS, Berriz G, Zhang L, Dupuy D, Walhout A, Cusick M, Roth F, Vidal M: Evidence for dynamically organized modularity in the yeast protein-protein interaction network. Nature. 2004, 430: 88-93. 10.1038/nature02555.
Rives A, Galitski T: Modular organisation of cellular networks. Proc Natl Acad Sci USA. 2003, 100: 1128-1133. 10.1073/pnas.0237338100.
Spirin V, Mirny L: Protein complexes and functional modules in molecular networks. Proc Natl Acad Sci USA. 2003, 100: 12123-12128. 10.1073/pnas.2032324100.
Wuchty S, Almaas E: Peeling the Yeast Interaction Network. Proteomics. 2005, 5: 444-449. 10.1002/pmic.200400962.
Snel B, Bork P, Huynen M: The identification of functional modules from genomic association of genes. Proc Natl Acad Sci USA. 2002, 99: 5890-5895. 10.1073/pnas.092632599.
Wuchty S: Interaction and Domain Networks of Yeast. Proteomics. 2002, 2: 1715-1723. 10.1002/1615-9861(200212)2:12<1715::AID-PROT1715>3.0.CO;2-O.
Wuchty S: Topology and Evolution in Yeast Interaction Networks. Genome Res. 2004, 14: 1310-1314. 10.1101/gr.2300204.
Fraser H, Hirsh A, Steinmetz L, Scharfe C, Feldman M: Evolutionary Rate in the Protein Interaction Network. Science. 2002, 296: 750-752. 10.1126/science.1068696.
Fraser H, Wall D, Hirsh A: A simple dependence between protein evolution rate and the number of protein-protein interactions. BMC Evol Biol. 2003, 3 (11):
Jordan I, Wolf Y, Koonin E: No simple dependence between protein evolution rate and the number of protein-protein interactions: only the most prolific interactors tend to evolve slowly. BMC Evol Biol. 2003, 3:
Jordan I, Wolf Y, Koonin E: Correction: No simple dependence between protein evolution rate and the number of protein-protein interactions: only the most prolific interactors tend to evolve slowly. BMC Evol Biol. 2003, 3 (5):
Williams E, Hurst L: The evolution of linked genes evolve at similar rates. Nature. 2000, 407: 900-902. 10.1038/35038066.
Wuchty S, Oltvai Z, Barabási AL: Evolutionary conservation of motif constituents within the yeast protein interaction network. Nature Genetics. 2003, 35: 176-179. 10.1038/ng1242.
Vespignani A: Evolution thinks modular. Nature Gen. 2003, 35: 118-119. 10.1038/ng1003-118.
von Mering C, Zdobnov E, Tsoka S, Ciccarelli F, JB Pereira-Leal CO, Bork P: Genome evolution reveals biochemical networks and functional modules. Proc Natl Acad Sci USA. 2003, 100: 15428-15433. 10.1073/pnas.2136809100.
Ge H, Liu Z, Church G, Vidal M: Correlation between transcriptome and interactome mapping data from Saccharomyces cerevisiae. Nature Genetics. 2001, 29: 482-486. 10.1038/ng776.
Babu M, Luscombe N, Aravind L, Gerstein M, Teichmann S: Structure and evolution of transcriptional regulatory networks. Curr Opin Struct Biol. 2004, 14: 283-291. 10.1016/j.sbi.2004.05.004.
Sharan R, Suthram S, Kelley R, Kuhn T, McCuine S, Uetz P, Sittler T, Karp R, Ideker T: Conserved patterns of protein interaction in multiple species. Proc Natl Acad Scie USA. 2005, 102: 1974-1979. 10.1073/pnas.0409522102.
Vidal M: Interactome modelling. FEBS Lett. 2005, 579: 1834-1838. 10.1016/j.febslet.2005.02.030.
Bork P, Jensen L, von Mering C, Ramani A, Lee I, Marcotte E: Protein interaction networks from yeast to human. Curr Opin Struct Biol. 2004, 14: 292-299. 10.1016/j.sbi.2004.05.003.
Vazquez A, Flammini A, Maritan A, Vespignani A: Modeling of Protein Interaction Networks. ComPlexUs. 2003, 1 (38): 38-44.
Von Mering C, Krause R, Snel B, Cornell M, Oliver S, Fields S, Bork P: Comparative assessment of large-scale data sets of protein-protein interactions. Nature. 2003, 417: 399-403.
Xenarios I, Salwinski L, Duan X, Higney P, Kim SM, Eisenberg D: DIP, the Database of Interacting Proteins: a research tool for studying cellular networks of protein interactions. Nucl Acids Res. 2002, 30: 303-305. 10.1093/nar/30.1.303.
Remm M, Storm C, Sonnhammer E: Automatic clustering of orthologs and in-paralogs from pairwise species comparisons. J Mol Biol. 2001, 314: 1041-1052. 10.1006/jmbi.2000.5197.
Goldberg D, Roth F: Assessing experimentally derived interactions in a small world. Proc Natl Acad Sci USA. 2003, 100: 4372-4376. 10.1073/pnas.0735871100.
Bozdech Z, Llinas M, Pulliam B, Wong E, Zhu J, DeRisi J: The Transcriptome of the Intraerythrocytic Developmental Cycle of Plasmodium falciparum. PLoS Biology. 2003, 1: 1-16. 10.1371/journal.pbio.0000005.
Eisen M, Spellman P, Brown P, Botstein D: Cluster analysis and display of genome-wide expression patterns. Proc Nat Acad Sci. 1998, 95: 14863-14868. 10.1073/pnas.95.25.14863.
Albert R, Jeong H, Barabási A: Error and attack tolerance of complex networks. Nature. 2000, 406: 378-382. 10.1038/35019019.
O'Brien K, Remm M, Sonnhammer E: Inparanoid: a comprehensive database of eukaryotic orthologs. Nucl Acids Res. 2005, 33: D476-D480. 10.1093/nar/gki107.
Goldstein M, Morris S, Yen G: Fitting to the Power-Law Distribution. 2004, [http://arxiv.org/abs/cond-mat/0402322]
M.T.F. is supported by NIH grant AI055025. A.-L.B. is supported by grants of NSF and NIH.