Genome evolution reveals biochemical networks and functional modules

Discovering functional linkages and uncharacterized cellular pathways using phylogenetic profile comparisons: a comprehensive assessment

BACKGROUND

A widely-used approach for discovering functional and physical interactions among proteins involves phylogenetic profile comparisons (PPCs). Here, proteins with similar profiles are inferred to be functionally related under the assumption that proteins involved in the same metabolic pathway or cellular system are likely to have been co-inherited during evolution.

RESULTS

Our experimentation with E. coli and yeast proteins with 16 different carefully composed reference sets of genomes revealed that the phyletic patterns of proteins in prokaryotes alone could be adequate enough to make reasonably accurate functional linkage predictions. A slight improvement in performance is observed on adding few eukaryotes into the reference set, but a noticeable drop-off in performance is observed with increased number of eukaryotes. Inclusion of most parasitic, pathogenic or vertebrate genomes and multiple strains of the same species into the reference set do not necessarily contribute to an improved sensitivity or accuracy. Interestingly, we also found that evolutionary histories of individual pathways have a significant affect on the performance of the PPC approach with respect to a particular reference set. For example, to accurately predict functional links in carbohydrate or lipid metabolism, a reference set solely composed of prokaryotic (or bacterial) genomes performed among the best compared to one composed of genomes from all three super-kingdoms; this is in contrast to predicting functional links in translation for which a reference set composed of prokaryotic (or bacterial) genomes performed the worst. We also demonstrate that the widely used random null model to quantify the statistical significance of profile similarity is incomplete, which could result in an increased number of false-positives.

CONCLUSION

Contrary to previous proposals, it is not merely the number of genomes but a careful selection of informative genomes in the reference set that influences the prediction accuracy of the PPC approach. We note that the predictive power of the PPC approach, especially in eukaryotes, is heavily influenced by the primary endosymbiosis and subsequent bacterial contributions. The over-representation of parasitic unicellular eukaryotes and vertebrates additionally make eukaryotes less useful in the reference sets. Reference sets composed of highly non-redundant set of genomes from all three super-kingdoms fare better with pathways showing considerable vertical inheritance and strong conservation (e.g. translation apparatus), while reference sets solely composed of prokaryotic genomes fare better for more variable pathways like carbohydrate metabolism. Differential performance of the PPC approach on various pathways, and a weak positive correlation between functional and profile similarities suggest that caution should be exercised while interpreting functional linkages inferred from genome-wide large-scale profile comparisons using a single reference set.

Distinctive topologies of partner-switching signaling networks correlate with their physiological roles

Regulatory networks controlling bacterial gene expression often evolve from common origins and share homologous proteins and similar network motifs. However, when functioning in different physiological contexts, these motifs may be re-arranged with different topologies that significantly affect network performance. Here we analyze two related signaling networks in the bacterium Bacillus subtilis in order to assess the consequences of their different topologies, with the aim of formulating design principles applicable to other systems. These two networks control the activities of the general stress response factor sigma(B) and the first sporulation-specific factor sigma(F). Both networks have at their core a "partner-switching" mechanism, in which an anti-sigma factor forms alternate complexes either with the sigma factor, holding it inactive, or with an anti-anti-sigma factor, thereby freeing sigma. However, clear differences in network structure are apparent: the anti-sigma factor for sigma(F) forms a long-lived, "dead-end" complex with its anti-anti-sigma factor and ADP, whereas the genes encoding sigma(B) and its network partners lie in a sigma(B)-controlled operon, resulting in positive and negative feedback loops. We constructed mathematical models of both networks and examined which features were critical for the performance of each design. The sigma(F) model predicts that the self-enhancing formation of the dead-end complex transforms the network into a largely irreversible hysteretic switch; the simulations reported here also demonstrate that hysteresis and slow turn off kinetics are the only two system properties associated with this complex formation. By contrast, the sigma(B) model predicts that the positive and negative feedback loops produce graded, reversible behavior with high regulatory capacity and fast response time. Our models demonstrate how alterations in network design result in different system properties that correlate with regulatory demands. These design principles agree with the known or suspected roles of similar networks in diverse bacteria.

Use of pathway analysis and genome context methods for functional genomics of Mycoplasma pneumoniae nucleotide metabolism

Elementary modes analysis allows one to reveal whether a set of known enzymes is sufficient to sustain functionality of the cell. Moreover, it is helpful in detecting missing reactions and predicting which enzymes could fill these gaps. Here, we perform a comprehensive elementary modes analysis and a genomic context analysis of Mycoplasma pneumoniae nucleotide metabolism, and search for new enzyme activities. The purine and pyrimidine networks are reconstructed by assembling enzymes annotated in the genome or found experimentally. We show that these reaction sets are sufficient for enabling synthesis of DNA and RNA in M. pneumoniae. Special focus is on the key modes for growth. Moreover, we make an educated guess on the nutritional requirements of this micro-organism. For the case that M. pneumoniae does not require adenine as a substrate, we suggest adenylosuccinate synthetase (EC 6.3.4.4), adenylosuccinate lyase (EC 4.3.2.2) and GMP reductase (EC 1.7.1.7) to be operative. GMP reductase activity is putatively assigned to the NRDI_MYCPN gene on the basis of the genomic context analysis. For the pyrimidine network, we suggest CTP synthase (EC 6.3.4.2) to be active. Further experiments on the nutritional requirements are needed to make a decision. Pyrimidine metabolism appears to be more appropriate as a drug target than purine metabolism since it shows lower plasticity.

The SsrA-SmpB ribosome rescue system is important for growth of Bacillus subtilis at low and high temperatures

Bacillus subtilis has multiple stress response systems whose integrated action promotes growth and survival under unfavorable conditions. Here we address the function and transcriptional organization of a five-gene cluster containing ssrA, previously known to be important for growth at high temperature because of the role of its tmRNA product in rescuing stalled ribosomes. Reverse transcription-PCR experiments detected a single message for the secG-yvaK-rnr-smpB-ssrA cluster, suggesting that it constitutes an operon. However, rapid amplification of cDNA ends-PCR and lacZ fusion experiments indicated that operon transcription is complex, with at least five promoters controlling different segments of the cluster. One sigma(A)-like promoter preceded secG (P(1)), and internal sigma(A)-like promoters were found in both the rnr-smpB (P(2)) and smpB-ssrA intervals (P(3) and P(HS)). Another internal promoter lay in the secG-yvaK intercistronic region, and this activity (P(B)) was dependent on the general stress factor sigma(B). Null mutations in the four genes downstream from P(B) were tested for their effects on growth. Loss of yvaK (carboxylesterase E) or rnr (RNase R) caused no obvious phenotype. By contrast, smpB was required for growth at high temperature (52 degrees C), as anticipated if its product (a small ribosomal binding protein) is essential for tmRNA (ssrA) function. Notably, smpB and ssrA were also required for growth at low temperature (16 degrees C), a phenotype not previously associated with tmRNA activity. These results extend the known high-temperature role of ssrA and indicate that the ribosome rescue system is important at both extremes of the B. subtilis temperature range.

A subsystems-based approach to the identification of drug targets in bacterial pathogens

This chapter describes a three-stage approach to target identification based upon subsystem analysis. Subsystems analysis focuses on related metabolic pathways as a unit and is a biochemically-informed approach to target selection. The process involves three stages of analysis; the first stage, selection of the target subsystem, is guided by information about its essentiality and on the predicted vulnerability of the targeted pathway or enzyme to inhibition. The second stage involves analysis of the target subsystem by means of comparative genomics, including genome context analysis and metabolic reconstruction. The third stage evaluates the selection of the specific target genes within the subsystem by target prioritization and validation. The whole process allows for a careful consideration of spectrum, drugability, biological rationale and the metabolic role of the specific target within the context of an integrated circuit within a specific metabolic pathway.

Stratus not altocumulus: a new view of the yeast protein interaction network

Systems biology approaches can reveal intermediary levels of organization between genotype and phenotype that often underlie biological phenomena such as polygenic effects and protein dispensability. An important conceptualization is the module, which is loosely defined as a cohort of proteins that perform a dedicated cellular task. Based on a computational analysis of limited interaction datasets in the budding yeast Saccharomyces cerevisiae, it has been suggested that the global protein interaction network is segregated such that highly connected proteins, called hubs, tend not to link to each other. Moreover, it has been suggested that hubs fall into two distinct classes: “party” hubs are co-expressed and co-localized with their partners, whereas “date” hubs interact with incoherently expressed and diversely localized partners, and thereby cohere disparate parts of the global network. This structure may be compared with altocumulus clouds, i.e., cotton ball–like structures sparsely connected by thin wisps. However, this organization might reflect a small and/or biased sample set of interactions. In a multi-validated high-confidence (HC) interaction network, assembled from all extant S. cerevisiae interaction data, including recently available proteome-wide interaction data and a large set of reliable literature-derived interactions, we find that hub–hub interactions are not suppressed. In fact, the number of interactions a hub has with other hubs is a good predictor of whether a hub protein is essential or not. We find that date hubs are neither required for network tolerance to node deletion, nor do date hubs have distinct biological attributes compared to other hubs. Date and party hubs do not, for example, evolve at different rates. Our analysis suggests that the organization of global protein interaction network is highly interconnected and hence interdependent, more like the continuous dense aggregations of stratus clouds than the segregated configuration of altocumulus clouds. If the network is configured in a stratus format, cross-talk between proteins is potentially a major source of noise. In turn, control of the activity of the most highly connected proteins may be vital. Indeed, we find that a fluctuation in steady-state levels of the most connected proteins is minimized.

Analysis of multi-validated protein interaction data reveals networks with greater interconnectivity than the more segregated structures seen in previously available data. To help visualize this, the authors draw comparisons between continuous stratus clouds and altocumulus clouds.

Disrupted in Schizophrenia 1 Interactome: evidence for the close connectivity of risk genes and a potential synaptic basis for schizophrenia

Disrupted in Schizophrenia 1 (DISC1) is a schizophrenia risk gene associated with cognitive deficits in both schizophrenics and the normal ageing population. In this study, we have generated a network of protein-protein interactions (PPIs) around DISC1. This has been achieved by utilising iterative yeast-two hybrid (Y2H) screens, combined with detailed pathway and functional analysis. This so-called 'DISC1 interactome' contains many novel PPIs and provides a molecular framework to explore the function of DISC1. The network implicates DISC1 in processes of cytoskeletal stability and organisation, intracellular transport and cell-cycle/division. In particular, DISC1 looks to have a PPI profile consistent with that of an essential synaptic protein, which fits well with the underlying molecular pathology observed at the synaptic level and the cognitive deficits seen behaviourally in schizophrenics. Utilising a similar approach with dysbindin (DTNBP1), a second schizophrenia risk gene, we show that dysbindin and DISC1 share common PPIs suggesting they may affect common biological processes and that the function of schizophrenia risk genes may converge.

Evolutionary-conserved gene expression response profiles across mammalian tissues

Gene expression responses are complex and frequently involve the actions of many genes to effect coordinated patterns. We hypothesized these coordinated responses are evolutionarily conserved and used a comparison of human and mouse gene expression profiles to identify the most prominent conserved features across a set of normal mammalian tissues. Based on data from multiple studies across multiple tissues in human and mouse, 13 gene expression modes across multiple tissues were identified in each of these species using principal component analysis. Strikingly, 1-to-1 pairing of human and mouse modes was observed in 12 out of 13 modes obtained from the two species independently. These paired modes define evolutionarily conserved gene expression response modes (CGEMs). Notably, in this study we were able to extract biological responses that are not overwhelmed by laboratory-to-laboratory or species-to-species variation. Of the variation in our gene expression dataset, 84% can be explained using these CGEMs. Functional annotation was performed using Gene Ontology, pathway, and transcription factor binding site over representation. Our conclusion is that we found an unbiased way of obtaining conserved gene response modes that accounts for a considerable portion of gene expression variation in a given dataset, as well as validates the conservation of major gene expression response modes across the mammals.

Modeling for evolving biological networks with scale-free connectivity, hierarchical modularity, and disassortativity

We propose a growing network model that consists of two tunable mechanisms: growth by merging modules which are represented as complete graphs and a fitness-driven preferential attachment. Our model exhibits the three prominent statistical properties are widely shared in real biological networks, for example gene regulatory, protein-protein interaction, and metabolic networks. They retain three power law relationships, such as the power laws of degree distribution, clustering spectrum, and degree-degree correlation corresponding to scale-free connectivity, hierarchical modularity, and disassortativity, respectively. After making comparisons of these properties between model networks and biological networks, we confirmed that our model has inference potential for evolutionary processes of biological networks.

Constrained models of evolution lead to improved prediction of functional linkage from correlated gain and loss of genes

MOTIVATION

We compare phylogenetic approaches for inferring functional gene links. The approaches detect independent instances of the correlated gain and loss of pairs of genes from species' genomes. We investigate the effect on results of basing evidence of correlations on two phylogenetic approaches, Dollo parsminony and maximum likelihood (ML). We further examine the effect of constraining the ML model by fixing the rate of gene gain at a low value, rather than estimating it from the data.

RESULTS

We detect correlated evolution among a test set of pairs of yeast (Saccharomyces cerevisiae) genes, with a case study of 21 eukaryotic genomes and test data derived from known yeast protein complexes. If the rate at which genes are gained is constrained to be low, ML achieves by far the best results at detecting known functional links. The model then has fewer parameters but it is more realistic by preventing genes from being gained more than once.

AVAILABILITY

BayesTraits by M. Pagel and A. Meade, and a script to configure and repeatedly launch it by D. Barker and M. Pagel, are available at http://www.evolution.reading.ac.uk

Coevolution, modularity and human disease

The concepts of coevolution and modularity have been studied separately for decades. Recent advances in genomics have led to the first systematic studies in each of these fields at the molecular level, resulting in several important discoveries. Both coevolution and modularity appear to be pervasive features of genomic data from all species studied to date, and their presence can be detected in many types of datasets, including genome sequences, gene expression data, and protein-protein interaction data. Moreover, the combination of these two ideas might have implications for our understanding of many aspects of biology, ranging from the general architecture of living systems to the causes of various human diseases.

LC-MS/MS based proteomic analysis and functional inference of hypothetical proteins in Desulfovibrio vulgaris

High efficiency capillary liquid chromatography-tandem mass spectrometry (LC-MS/MS) was used to examine the proteins extracted from Desulfovibrio vulgaris cells across six treatment conditions. While our previous study provided a proteomic overview of the cellular metabolism based on proteins with known functions [W. Zhang, M.A. Gritsenko, R.J. Moore, D.E. Culley, L. Nie, K. Petritis, E.F. Strittmatter, D.G. Camp II, R.D. Smith, F.J. Brockman, A proteomic view of the metabolism in Desulfovibrio vulgaris determined by liquid chromatography coupled with tandem mass spectrometry, Proteomics 6 (2006) 4286-4299], this study describes the global detection and functional inference for hypothetical D. vulgaris proteins. Using criteria that a given peptide of a protein is identified from at least two out of three independent LC-MS/MS measurements and that for any protein at least two different peptides are identified among the three measurements, 129 open reading frames (ORFs) originally annotated as hypothetical proteins were found to encode expressed proteins. Functional inference for the conserved hypothetical proteins was performed by a combination of several non-homology based methods: genomic context analysis, phylogenomic profiling, and analysis of a combination of experimental information, including peptide detection in cells grown under specific culture conditions and cellular location of the proteins. Using this approach we were able to assign possible functions to 20 conserved hypothetical proteins. This study demonstrated that a combination of proteomics and bioinformatics methodologies can provide verification of the expression of hypothetical proteins and improve genome annotation.

Essential genes on metabolic maps

Within the past five years genome-scale gene essentiality data sets have been published for ten diverse bacterial species. These data are a rich source of information about cellular networks that we are only beginning to explore. The analysis of these data, very heterogeneous in nature, is a challenging task. Even the definition of 'essential genes' in various genome-scale studies varies from genes 'absolutely required for survival' to those 'strongly contributing to fitness' and robust competitive growth. A comparative analysis of gene essentiality across multiple organisms based on projection of experimentally observed essential genes to functional roles in a collection of metabolic pathways and subsystems is emerging as a powerful tool of systems biology.

Gene-balanced duplications, like tetraploidy, provide predictable drive to increase morphological complexity

Controversy surrounds the apparent rising maximums of morphological complexity during eukaryotic evolution, with organisms increasing the number and nestedness of developmental areas as evidenced by morphological elaborations reflecting area boundaries. No "predictable drive" to increase this sort of complexity has been reported. Recent genetic data and theory in the general area of gene dosage effects has engendered a robust "gene balance hypothesis," with a theoretical base that makes specific predictions as to gene content changes following different types of gene duplication. Genomic data from both chordate and angiosperm genomes fit these predictions: Each type of duplication provides a one-way injection of a biased set of genes into the gene pool. Tetraploidies and balanced segments inject bias for those genes whose products are the subunits of the most complex biological machines or cascades, like transcription factors (TFs) and proteasome core proteins. Most duplicate genes are removed after tetraploidy. Genic balance is maintained by not removing those genes that are dose-sensitive, which tends to leave duplicate "functional modules" as the indirect products (spandrels) of purifying selection. Functional modules are the likely precursors of coadapted gene complexes, a unit of natural selection. The result is a predictable drive mechanism where "drive" is used rigorously, as in "meiotic drive." Rising morphological gain is expected given a supply of duplicate functional modules. All flowering plants have survived at least three large-scale duplications/diploidizations over the last 300 million years (Myr). An equivalent period of tetraploidy and body plan evolution may have ended for animals 500 million years ago (Mya). We argue that "balanced gene drive" is a sufficient explanation for the trend that the maximums of morphological complexity have gone up, and not down, in both plant and animal eukaryotic lineages.

The outcomes of pathway database computations depend on pathway ontology

Different biological notions of pathways are used in different pathway databases. Those pathway ontologies significantly impact pathway computations. Computational users of pathway databases will obtain different results depending on the pathway ontology used by the databases they employ, and different pathway ontologies are preferable for different end uses. We explore differences in pathway ontologies by comparing the BioCyc and KEGG ontologies. The BioCyc ontology defines a pathway as a conserved, atomic module of the metabolic network of a single organism, i.e. often regulated as a unit, whose boundaries are defined at high-connectivity stable metabolites. KEGG pathways are on average 4.2 times larger than BioCyc pathways, and combine multiple biological processes from different organisms to produce a substrate-centered reaction mosaic. We compared KEGG and BioCyc pathways using genome context methods, which determine the functional relatedness of pairs of genes. For each method we employed, a pair of genes randomly selected from a BioCyc pathway is more likely to be related by that method than is a pair of genes randomly selected from a KEGG pathway, supporting the conclusion that the BioCyc pathway conceptualization is closer to a single conserved biological process than is that of KEGG.

Maps, books and other metaphors for systems biology

We briefly review the use of metaphors in science and progressively focus on fields from biology and molecular biology to genomics and bioinformatics. We discuss how metaphors are both a tool for scientific exploration and a medium for public communication of complex subjects, by various short examples. Finally, we propose a metaphor for systems biology that provides an illuminating perspective for the ambitious goals of this field and delimits its current agenda.

Accelerating the reconstruction of genome-scale metabolic networks

BACKGROUND

The genomic information of a species allows for the genome-scale reconstruction of its metabolic capacity. Such a metabolic reconstruction gives support to metabolic engineering, but also to integrative bioinformatics and visualization. Sequence-based automatic reconstructions require extensive manual curation, which can be very time-consuming. Therefore, we present a method to accelerate the time-consuming process of network reconstruction for a query species. The method exploits the availability of well-curated metabolic networks and uses high-resolution predictions of gene equivalency between species, allowing the transfer of gene-reaction associations from curated networks.

RESULTS

We have evaluated the method using Lactococcus lactis IL1403, for which a genome-scale metabolic network was published recently. We recovered most of the gene-reaction associations (i.e. 74 - 85%) which are incorporated in the published network. Moreover, we predicted over 200 additional genes to be associated to reactions, including genes with unknown function, genes for transporters and genes with specific metabolic reactions, which are good candidates for an extension to the previously published network. In a comparison of our developed method with the well-established approach Pathologic, we predicted 186 additional genes to be associated to reactions. We also predicted a relatively high number of complete conserved protein complexes, which are derived from curated metabolic networks, illustrating the potential predictive power of our method for protein complexes.

CONCLUSION

We show that our methodology can be applied to accelerate the reconstruction of genome-scale metabolic networks by taking optimal advantage of existing, manually curated networks. As orthology detection is the first step in the method, only the translated open reading frames (ORFs) of a newly sequenced genome are necessary to reconstruct a metabolic network. When more manually curated metabolic networks will become available in the near future, the usefulness of our method in network prediction is likely to increase.

A metabolic network in the evolutionary context: multiscale structure and modularity

The enormous complexity of biological networks has led to the suggestion that networks are built of modules that perform particular functions and are "reused" in evolution in a manner similar to reusable domains in protein structures or modules of electronic circuits. Analysis of known biological networks has revealed several modules, many of which have transparent biological functions. However, it remains to be shown that identified structural modules constitute evolutionary building blocks, independent and easily interchangeable units. An alternative possibility is that evolutionary modules do not match structural modules. To investigate the structure of evolutionary modules and their relationship to functional ones, we integrated a metabolic network with evolutionary associations between genes inferred from comparative genomics. The resulting metabolic-genomic network places metabolic pathways into evolutionary and genomic context, thereby revealing previously unknown components and modules. We analyzed the integrated metabolic-genomic network on three levels: macro-, meso-, and microscale. The macroscale level demonstrates strong associations between neighboring enzymes and between enzymes that are distant on the network but belong to the same linear pathway. At the mesoscale level, we identified evolutionary metabolic modules and compared them with traditional metabolic pathways. Although, in some cases, there is almost exact correspondence, some pathways are split into independent modules. On the microscale level, we observed high association of enzyme subunits and weak association of isoenzymes independently catalyzing the same reaction. This study shows that evolutionary modules, rather than pathways, may be thought of as regulatory and functional units in bacterial genomes.

The origins and evolution of functional modules: lessons from protein complexes

Modularity is an attribute of a system that can be decomposed into a set of cohesive entities that are loosely coupled. Many cellular networks can be decomposed into functional modules-each functionally separable from the other modules. The protein complexes in physical protein interaction networks are a good example of this, and here we focus on their origins and evolution. We investigate the emergence of protein complexes and physical interactions between proteins by duplication, and review other mechanisms. We dissect the dataset of protein complexes of known three-dimensional structure, and show that roughly 90% of these complexes contain contacts between identical proteins within the same complex. Proteins that are shared across different complexes occur frequently, and they tend to be essential genes more often than members of a single protein complex. We also provide a perspective on the evolutionary mechanisms driving the growth of other modular cellular networks such as transcriptional regulatory and metabolic networks.

Multi-target strategies for the improved treatment of depressive states: Conceptual foundations and neuronal substrates, drug discovery and therapeutic application

Major depression is a debilitating and recurrent disorder with a substantial lifetime risk and a high social cost. Depressed patients generally display co-morbid symptoms, and depression frequently accompanies other serious disorders. Currently available drugs display limited efficacy and a pronounced delay to onset of action, and all provoke distressing side effects. Cloning of the human genome has fuelled expectations that symptomatic treatment may soon become more rapid and effective, and that depressive states may ultimately be "prevented" or "cured". In pursuing these objectives, in particular for genome-derived, non-monoaminergic targets, "specificity" of drug actions is often emphasized. That is, priority is afforded to agents that interact exclusively with a single site hypothesized as critically involved in the pathogenesis and/or control of depression. Certain highly selective drugs may prove effective, and they remain indispensable in the experimental (and clinical) evaluation of the significance of novel mechanisms. However, by analogy to other multifactorial disorders, "multi-target" agents may be better adapted to the improved treatment of depressive states. Support for this contention is garnered from a broad palette of observations, ranging from mechanisms of action of adjunctive drug combinations and electroconvulsive therapy to "network theory" analysis of the etiology and management of depressive states. The review also outlines opportunities to be exploited, and challenges to be addressed, in the discovery and characterization of drugs recognizing multiple targets. Finally, a diversity of multi-target strategies is proposed for the more efficacious and rapid control of core and co-morbid symptoms of depression, together with improved tolerance relative to currently available agents.

Prediction of yeast protein-protein interaction network: insights from the Gene Ontology and annotations

A map of protein-protein interactions provides valuable insight into the cellular function and machinery of a proteome. By measuring the similarity between two Gene Ontology (GO) terms with a relative specificity semantic relation, here, we proposed a new method of reconstructing a yeast protein-protein interaction map that is solely based on the GO annotations. The method was validated using high-quality interaction datasets for its effectiveness. Based on a Z-score analysis, a positive dataset and a negative dataset for protein-protein interactions were derived. Moreover, a gold standard positive (GSP) dataset with the highest level of confidence that covered 78% of the high-quality interaction dataset and a gold standard negative (GSN) dataset with the lowest level of confidence were derived. In addition, we assessed four high-throughput experimental interaction datasets using the positives and the negatives as well as GSPs and GSNs. Our predicted network reconstructed from GSPs consists of 40,753 interactions among 2259 proteins, and forms 16 connected components. We mapped all of the MIPS complexes except for homodimers onto the predicted network. As a result, approximately 35% of complexes were identified interconnected. For seven complexes, we also identified some nonmember proteins that may be functionally related to the complexes concerned. This analysis is expected to provide a new approach for predicting the protein-protein interaction maps from other completely sequenced genomes with high-quality GO-based annotations.

Locally defined protein phylogenetic profiles reveal previously missed protein interactions and functional relationships

Phylogenetic profiles encode patterns of presence or absence of genes across genomes, and these profiles can be used to assign functional relationships to nonhomologous pairs of proteins (Pellegrini et al., Proc Natl Acad Sci USA 1999;96:4284-4288). Although it is well known that many proteins were created from combinations of domains, most of the existing implementations of phylogenetic profiles do not consider this fact. Here, we introduce an extension that considers the multidomain nature of proteins and test the method against the known interaction data sets. Whereas earlier implementations associated one entire sequence with one protein phylogenetic profile (Single-Profile), our method instead breaks the sequence into a set of segments of predetermined size and constructs a separate profile for each segment (Multiple-Profile). The results show that the Multiple-Profile method performs as well as the Single-Profile method. However, the two methods share, surprisingly, a small fraction of their predictions, indicating that the Multiple-Profile method can detect known interactions missed by the Single-Profile method. Thus, the Multiple-Profile method can be used with other methods to determine functional relationships on a genome scale with wider coverage.

Identifying metabolic enzymes with multiple types of association evidence

Background

Existing large-scale metabolic models of sequenced organisms commonly include enzymatic functions which can not be attributed to any gene in that organism. Existing computational strategies for identifying such missing genes rely primarily on sequence homology to known enzyme-encoding genes.

Results

We present a novel method for identifying genes encoding for a specific metabolic function based on a local structure of metabolic network and multiple types of functional association evidence, including clustering of genes on the chromosome, similarity of phylogenetic profiles, gene expression, protein fusion events and others. Using E. coli and S. cerevisiae metabolic networks, we illustrate predictive ability of each individual type of association evidence and show that significantly better predictions can be obtained based on the combination of all data. In this way our method is able to predict 60% of enzyme-encoding genes of E. coli metabolism within the top 10 (out of 3551) candidates for their enzymatic function, and as a top candidate within 43% of the cases.

Conclusion

We illustrate that a combination of genome context and other functional association evidence is effective in predicting genes encoding metabolic enzymes. Our approach does not rely on direct sequence homology to known enzyme-encoding genes, and can be used in conjunction with traditional homology-based metabolic reconstruction methods. The method can also be used to target orphan metabolic activities.

Extraction of phylogenetic network modules from the metabolic network

Background

In bio-systems, genes, proteins and compounds are related to each other, thus forming complex networks. Although each organism has its individual network, some organisms contain common sub-networks based on function. Given a certain sub-network, the distribution of organisms common to it represents the diversity of its function.

Results

We extracted such "common" sub-networks, defined as "phylogenetic network modules," using phylogenetic profiles and cluster analysis. The enzymes in the same "phylogenetic network module" have similar phylogenetic profiles and related functions. These modules are shown to be phylogenetic building blocks. Furthermore, the network of the modules illustrated hierarchical feature as well as the network of enzymes involved in the metabolism.

Conclusion

We conclude that phylogenetic network modules are evolutionary conserved functional units in the metabolic network. We claim that our concept of phylogenetic modules provides a more accurate understanding of the evolution of biological networks.

Gene annotation and network inference by phylogenetic profiling

BACKGROUND

Phylogenetic analysis is emerging as one of the most informative computational methods for the annotation of genes and identification of evolutionary modules of functionally related genes. The effectiveness with which phylogenetic profiles can be utilized to assign genes to pathways depends on an appropriate measure of correlation between gene profiles, and an effective decision rule to use the correlate. Current methods, though useful, perform at a level well below what is possible, largely because performance of the latter deteriorates rapidly as coverage increases.

RESULTS

We introduce, test and apply a new decision rule, correlation enrichment (CE), for assigning genes to functional categories at various levels of resolution. Among the results are: (1) CE performs better than standard guilt by association (SGA, assignment to a functional category when a simple correlate exceeds a pre-specified threshold) irrespective of the number of genes assigned (i.e. coverage); improvement is greatest at high coverage where precision (positive predictive value) of CE is approximately 6-fold higher than that of SGA. (2) CE is estimated to allocate each of the 2918 unannotated orthologs to KEGG pathways with an average precision of 49% (approximately 7-fold higher than SGA) (3) An estimated 94% of the 1846 unannotated orthologs in the COG ontology can be assigned a function with an average precision of 0.4 or greater. (4) Dozens of functional and evolutionarily conserved cliques or quasi-cliques can be identified, many having previously unannotated genes.

CONCLUSION

The method serves as a general computational tool for annotating large numbers of unknown genes, uncovering evolutionary and functional modules. It appears to perform substantially better than extant stand alone high throughout methods.

Predicting genes for orphan metabolic activities using phylogenetic profiles

Homology-based methods fail to assign genes to many metabolic activities present in sequenced organisms. To suggest genes for these orphan activities we developed a novel method that efficiently combines local structure of a metabolic network with phylogenetic profiles. We validated our method using known metabolic genes in Saccharomyces cerevisiae and Escherichia coli. We show that our method should be easily transferable to other organisms, and that it is robust to errors in incomplete metabolic networks.

Identification and analysis of evolutionarily cohesive functional modules in protein networks

The increasing number of sequenced genomes makes it possible to infer the evolutionary history of functional modules, i.e., groups of proteins that contribute jointly to the same cellular function in a given species. Here we identify and analyze those prokaryotic functional modules, whose composition remains largely unchanged during evolution, and study their properties. Such "cohesive" modules have a large number of internal functional connections, encode genes that tend to be in close proximity in prokaryotic genomes, and correspond to physical complexes or complex functional systems like the flagellar apparatus. Cohesive modules are enriched in processes such as energy and amino acid metabolism, cell motility, and intracellular trafficking, or secretion. By grouping genes into modules we achieve a more precise estimate of their age and find that the young modules are often horizontally transferred between species and are enriched in functions involved in interactions with the environment, implying that they play an important role in the adaptation of species to new environments.

Stable evolutionary signal in a yeast protein interaction network

Background

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.

Results

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.

Conclusion

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.

Inferring functional information from domain co-evolution

MOTIVATION

Co-evolution is a powerful mechanism for understanding protein function. Prior work in this area has shown that co-evolving proteins are more likely to share the same function than those that do not because of functional constraints. Many of the efforts founded on this observation, however, are at the level of entire sequences, implicitly assuming that the complete protein sequence follows a single evolutionary trajectory. Since it is well known that a domain can exist in various contexts, this assumption is not valid for numerous multi-domain proteins. Motivated by these observations, we introduce a novel technique called Coevolutionary-Matrix that captures co-evolution between regions of two proteins. Instead of using existing domain information, the method exploits residue-level conservation to identify co-evolving regions that might correspond to domains.

RESULTS

We show that the Coevolutionary-Matrix method can detect greater number of known functional associations for the Escherichia coli proteins when compared with earlier implementations of phylogenetic profiles. Furthermore, co-evolving regions of proteins detected by our method enable us to make hypotheses about their specific functions, many of which are supported by existing biochemical studies.

Finding instabilities in the community structure of complex networks

The problem of finding clusters in complex networks has been studied by mathematicians, computer scientists, and, more recently, by physicists. Many of the existing algorithms partition a network into clear clusters without overlap. Here we introduce a method to identify the nodes lying "between clusters," allowing for a general measure of the stability of the clusters. This is done by adding noise over the edge weights. Our method can in principle be used with almost any clustering algorithm able to deal with weighted networks. We present several applications on real-world networks using two different clustering algorithms.

Evolutionary, structural and functional relationships revealed by comparative analysis of syntenic genes in Rhizobiales

BACKGROUND

Comparative genomics has provided valuable insights into the nature of gene sequence variation and chromosomal organization of closely related bacterial species. However, questions about the biological significance of gene order conservation, or synteny, remain open. Moreover, few comprehensive studies have been reported for rhizobial genomes.

RESULTS

We analyzed the genomic sequences of four fast growing Rhizobiales (Sinorhizobium meliloti, Agrobacterium tumefaciens, Mesorhizobium loti and Brucella melitensis). We made a comprehensive gene classification to define chromosomal orthologs, genes with homologs in other replicons such as plasmids, and those which were species-specific. About two thousand genes were predicted to be orthologs in each chromosome and about 80% of these were syntenic. A striking gene colinearity was found in pairs of organisms and a large fraction of the microsyntenic regions and operons were similar. Syntenic products showed higher identity levels than non-syntenic ones, suggesting a resistance to sequence variation due to functional constraints; also, an unusually high fraction of syntenic products contained membranal segments. Syntenic genes encode a high proportion of essential cell functions, presented a high level of functional relationships and a very low horizontal gene transfer rate. The sequence variability of the proteins can be considered the species signature in response to specific niche adaptation. Comparatively, an analysis with genomes of Enterobacteriales showed a different gene organization but gave similar results in the synteny conservation, essential role of syntenic genes and higher functional linkage among the genes of the microsyntenic regions.

CONCLUSION

Syntenic bacterial genes represent a commonly evolved group. They not only reveal the core chromosomal segments present in the last common ancestor and determine the metabolic characteristics shared by these microorganisms, but also show resistance to sequence variation and rearrangement, possibly due to their essential character. In Rhizobiales and Enterobacteriales, syntenic genes encode a high proportion of essential cell functions and presented a high level of functional relationships.

The viriosphere, diversity, and genetic exchange within phage communities

Natural phage communities are reservoirs of the greatest uncharacterized genetic diversity on Earth. Yet, identical phage sequences can be found in extremely different environments, which implies that there is wide circulation of viral genes among distantly related host populations. Further evidence of genetic exchange among phage and host communities is the presence in phage of genes coding for proteins that are essential for photosynthesis. These observations support the idea that a primary role of host populations in phage ecology and evolution is to serve as vectors for genetic exchange.

From proteins to proteomics

During the second half of the 20th century, biochemistry and subsequently molecular biology blossomed into the core upon which all biological and biomedical sciences now depend. A major part of these closely related disciplines has been the study of the structure and function of proteins and the diverse biological functions that they perform. Early experimentation necessarily focused on individual entities, selected mainly for their activities, but as technology improved there developed a tendency to look at proteins as larger, interactive groups or clusters. Spurred by the recent exponential production of genomic sequence data for a rapidly increasing number of species, protein chemistry has now evolved into a new discipline, proteomics. In addition to embracing the methods and approaches that have served protein scientists well in the past, it includes, and is perhaps best defined by, high-throughput analyses based in large part on 2D gel electrophoresis, MALDI and ESI mass spectrometry and combinatorial arrays. Proteomic targets include the identification of all genome products and a mapping of their interactions and expression profiles. These hold great promise for the identification of disease markers and drug targets, but are not without their challenges and pitfalls.

Genome annotation errors in pathway databases due to semantic ambiguity in partial EC numbers

We report on a new type of systematic annotation error in genome and pathway databases that results from the misinterpretation of partial Enzyme Commission (EC) numbers such as ‘1.1.1.-’. This error results in the assignment of genes annotated with a partial EC number to many or all biochemical reactions that are annotated with the same partial EC number. That inference is faulty because of the ambiguous nature of partial EC numbers. We have observed this type of error in multiple databases, including KEGG, VIMSS and IMG, all of which assign genes to KEGG pathways. The Escherichia coli subset of the KEGG database exhibits this error for 6.8% of its gene-reaction assignments. For example, KEGG contains 17 reactions that are annotated with EC 1.1.1.-. A group of three E.coli genes, b1580 [putative dehydrogenase, NAD(P)-binding, starvation-sensing protein], b3787 (UDP-N-acetyl-d-mannosaminuronic acid dehydrogenase) and b0207 (2,5-diketo-d-gluconate reductase B), is assigned to 15 of those reactions, despite experimental evidence indicating different single functions for two of the three genes. Furthermore, the databases (DBs) are internally inconsistent in that the description of gene functions for genes with partial EC numbers is inconsistent with the activities implied by reactions to which the genes were assigned. We infer that these inconsistencies result from the processing used to match gene products to reactions within KEGG's metabolic pathways. These errors affect scientists who use these DBs as online encyclopedias and they affect bioinformaticists who use these DBs to train and validate newly developed algorithms.

Computational characterization of proteins

Computational characterization of proteins is a necessary first step in understanding the biologic role of a protein. The composite architecture of mammalian proteins makes the prediction of the biologic role rather difficult. Nevertheless, integration of many different prediction methods allows for a more accurate representation. Information on the 3D structure of a protein improves the reliability of predictions of many features. This article reviews existing methods used to characterize proteins and several tools that provide an integrated access to different types of information. The authors point out the increasing importance of structural constraints and an increasing need to integrate different approaches.

Prediction of protein-protein interactions using distant conservation of sequence patterns and structure relationships

MOTIVATION

Given that association and dissociation of protein molecules is crucial in most biological processes several in silico methods have been recently developed to predict protein-protein interactions. Structural evidence has shown that usually interacting pairs of close homologs (interologs) physically interact in the same way. Moreover, conservation of an interaction depends on the conservation of the interface between interacting partners. In this article we make use of both, structural similarities among domains of known interacting proteins found in the Database of Interacting Proteins (DIP) and conservation of pairs of sequence patches involved in protein-protein interfaces to predict putative protein interaction pairs.

RESULTS

We have obtained a large amount of putative protein-protein interaction (approximately 130,000). The list is independent from other techniques both experimental and theoretical. We separated the list of predictions into three sets according to their relationship with known interacting proteins found in DIP. For each set, only a small fraction of the predicted protein pairs could be independently validated by cross checking with the Human Protein Reference Database (HPRD). The fraction of validated protein pairs was always larger than that expected by using random protein pairs. Furthermore, a correlation map of interacting protein pairs was calculated with respect to molecular function, as defined in the Gene Ontology database. It shows good consistency of the predicted interactions with data in the HPRD database. The intersection between the lists of interactions of other methods and ours produces a network of potentially high-confidence interactions.

Predicting functional gene links from phylogenetic-statistical analyses of whole genomes

An important element of the developing field of proteomics is to understand protein-protein interactions and other functional links amongst genes. Across-species correlation methods for detecting functional links work on the premise that functionally linked proteins will tend to show a common pattern of presence and absence across a range of genomes. We describe a maximum likelihood statistical model for predicting functional gene linkages. The method detects independent instances of the correlated gain or loss of pairs of proteins on phylogenetic trees, reducing the high rates of false positives observed in conventional across-species methods that do not explicitly incorporate a phylogeny. We show, in a dataset of 10,551 protein pairs, that the phylogenetic method improves by up to 35% on across-species analyses at identifying known functionally linked proteins. The method shows that protein pairs with at least two to three correlated events of gain or loss are almost certainly functionally linked. Contingent evolution, in which one gene's presence or absence depends upon the presence of another, can also be detected phylogenetically, and may identify genes whose functional significance depends upon its interaction with other genes. Incorporating phylogenetic information improves the prediction of functional linkages. The improvement derives from having a lower rate of false positives and from detecting trends that across-species analyses miss. Phylogenetic methods can easily be incorporated into the screening of large-scale bioinformatics datasets to identify sets of protein links and to characterise gene networks.

Prediction of functional modules based on comparative genome analysis and Gene Ontology application

We present a computational method for the prediction of functional modules encoded in microbial genomes. In this work, we have also developed a formal measure to quantify the degree of consistency between the predicted and the known modules, and have carried out statistical significance analysis of consistency measures. We first evaluate the functional relationship between two genes from three different perspectives--phylogenetic profile analysis, gene neighborhood analysis and Gene Ontology assignments. We then combine the three different sources of information in the framework of Bayesian inference, and we use the combined information to measure the strength of gene functional relationship. Finally, we apply a threshold-based method to predict functional modules. By applying this method to Escherichia coli K12, we have predicted 185 functional modules. Our predictions are highly consistent with the previously known functional modules in E.coli. The application results have demonstrated that our approach is highly promising for the prediction of functional modules encoded in a microbial genome.

Comparative metagenomics of microbial communities

The species complexity of microbial communities and challenges in culturing representative isolates make it difficult to obtain assembled genomes. Here we characterize and compare the metabolic capabilities of terrestrial and marine microbial communities using largely unassembled sequence data obtained by shotgun sequencing DNA isolated from the various environments. Quantitative gene content analysis reveals habitat-specific fingerprints that reflect known characteristics of the sampled environments. The identification of environment-specific genes through a gene-centric comparative analysis presents new opportunities for interpreting and diagnosing environments.

From protein networks to biological systems

A system-level understanding of any biological process requires a map of the relationships among the various molecules involved. Technologies to detect and predict protein interactions have begun to produce very large maps of protein interactions, some including most of an organism's proteins. These maps can be used to study how proteins work together to form molecular machines and regulatory pathways. They also provide a framework for constructing predictive models of how information and energy flow through biological networks. In many respects, protein interaction maps are an entrée into systems biology.

The evolution of Ca2+ signalling in photosynthetic eukaryotes

It is likely that cytosolic Ca2+ elevations have played a part in eukaryotic signal transduction for about the last 2 Gyr, being mediated by a group of molecules which are collectively known as the [Ca2+]cyt signalling toolkit. Different eukaryotes often display strikingly similar [Ca2+]cyt signalling elevations, which may reflect conservation of toolkit components (homology) or similar constraints acting on different toolkits (homoplasy). Certain toolkit components, which are presumably ancestral, are shared by plants and animals, but some components are unique to photosynthetic organisms. We propose that the structure of modern plant [Ca2+]cyt signalling toolkits may be explained by their modular adaptation from earlier pathways.

Analysis of genomic context: prediction of functional associations from conserved bidirectionally transcribed gene pairs

Several widely used methods for predicting functional associations between proteins are based on the systematic analysis of genomic context. Efforts are ongoing to improve these methods and to search for novel aspects in genomes that could be exploited for function prediction. Here, we use gene expression data to demonstrate two functional implications of genome organization: first, chromosomal proximity indicates gene coregulation in prokaryotes independent of relative gene orientation; and second, adjacent bidirectionally transcribed genes (that is,'divergently' organized coding regions) with conserved gene orientation are strongly coregulated. We further demonstrate that such bidirectionally transcribed gene pairs are functionally associated and derive from this a novel genomic context method that reliably predicts links between >2,500 pairs of genes in approximately 100 species. Around 650 of these functional associations are supported by other genomic context methods. In most instances, one gene encodes a transcriptional regulator, and the other a nonregulatory protein. In-depth analysis in Escherichia coli shows that the vast majority of these regulators both control transcription of the divergently transcribed target gene/operon and auto-regulate their own biosynthesis. The method thus enables the prediction of target processes and regulatory features for several hundred transcriptional regulators.

Protein interaction networks from yeast to human

Protein interaction networks summarize large amounts of protein-protein interaction data, both from individual, small-scale experiments and from automated high-throughput screens. The past year has seen a flood of new experimental data, especially on metazoans, as well as an increasing number of analyses designed to reveal aspects of network topology, modularity and evolution. As only minimal progress has been made in mapping the human proteome using high-throughput screens, the transfer of interaction information within and across species has become increasingly important. With more and more heterogeneous raw data becoming available, proper data integration and quality control have become essential for reliable protein network reconstruction, and will be especially important for reconstructing the human protein interaction network.

A probabilistic functional network of yeast genes

A conceptual framework for integrating diverse functional genomics data was developed by reinterpreting experiments to provide numerical likelihoods that genes are functionally linked. This allows direct comparison and integration of different classes of data. The resulting probabilistic gene network estimates the functional coupling between genes. Within this framework, we reconstructed an extensive, high-quality functional gene network for Saccharomyces cerevisiae, consisting of 4681 (approximately 81%) of the known yeast genes linked by approximately 34,000 probabilistic linkages comparable in accuracy to small-scale interaction assays. The integrated linkages distinguish true from false-positive interactions in earlier data sets; new interactions emerge from genes' network contexts, as shown for genes in chromatin modification and ribosome biogenesis.

Functional clues for hypothetical proteins based on genomic context analysis in prokaryotes

Three integrated genomic context methods were used to annotate uncharacterized proteins in 102 bacterial genomes. Of 7853 orthologous groups with unknown function containing 45,110 proteins, 1738 groups could be linked to functionally associated partners. In many cases, those partners are uncharacterized themselves (hinting at newly identified modules) or have been described in general terms only. However, we were able to assign pathways, cellular processes or physical complexes for 273 groups (encompassing 3624 previously functionally uncharacterized proteins).

Towards a comprehensive understanding ofBacillus subtiliscell physiology by physiological proteomics

Using Bacillus subtilis as a model system for functional genomics, this review will provide insights how proteomics can be used to bring the virtual life of genes to the real life of proteins. Physiological proteomics will generate a new and broad understanding of cellular physiology because the majority of proteins synthesized in the cell can be visualized. From a physiological point of view two major proteome fractions can be distinguished: proteomes of growing cells and proteomes of nongrowing cells. In the main analytical window almost 50% of the vegetative proteome expressed in growing cells of B. subtilis were identified. This proteomic view of growing cells can be employed for analyzing the regulation of entire metabolic pathways and thus opens the chance for a comprehensive understanding of metabolism and growth processes of bacteria. Proteomics, on the other hand, is also a useful tool for analyzing the adaptational network of nongrowing cells that consists of several partially overlapping regulation groups induced by stress/starvation stimuli. Furthermore, proteomic signatures for environmental stimuli can not only be applied to predict the physiological state of cells, but also offer various industrial applications from fermentation monitoring up to the analysis of the mode of action of drugs. Even if DNA array technologies currently provide a better overview of the gene expression profile than proteome approaches, the latter address biological problems in which they can not be replaced by mRNA profiling procedures. This proteomics of the second generation is a powerful tool for analyzing global control of protein stability, the protein interaction network, protein secretion or post-translational modifications of proteins on the way towards the elucidation of the mystery of life.

Development through the eyes of functional genomics

In many of the model organisms used to study development, it is becoming relatively routine to carry out global analyses of gene function. These analyses take many forms, from microarray analyses to the construction of physical interaction maps to the systematic analyses of loss-of-function phenotypes. Such large-scale datasets can be integrated to generate complex gene networks, and we explore how these gene networks can contribute to an understanding of developmental pathways. In particular, we examine how combining large-scale expression experiments and gene networks may move us towards a molecular description of the events of development, embodied in a succession of stage-specific subnetworks sampled from an organism's overall gene network.

A probabilistic view of gene function

Cells are controlled by the complex and dynamic actions of thousands of genes. With the sequencing of many genomes, the key problem has shifted from identifying genes to knowing what the genes do; we need a framework for expressing that knowledge. Even the most rigorous attempts to construct ontological frameworks describing gene function (e.g., the Gene Ontology project) ultimately rely on manual curation and are thus labor-intensive and subjective. But an alternative exists: the field of functional genomics is piecing together networks of gene interactions, and although these data are currently incomplete and error-prone, they provide a glimpse of a new, probabilistic view of gene function. We outline such a framework, which revolves around a statistical description of gene interactions derived from large, systematically compiled data sets. In this probabilistic view, pleiotropy is implicit, all data have errors and the definition of gene function is an iterative process that ultimately converges on the correct functions. The relationships between the genes are defined by the data, not by hand. Even this comprehensive view fails to capture key aspects of gene function, not least their dynamics in time and space, showing that there are limitations to the model that must ultimately be addressed.