Metabolic-network-driven analysis of bacterial ecological strategies
- Shiri Freilich†1, 2Email author,
- Anat Kreimer†3,
- Elhanan Borenstein4, 5,
- Nir Yosef1,
- Roded Sharan1,
- Uri Gophna6 and
- Eytan Ruppin1, 2Email author
© Freilich et al.; licensee BioMed Central Ltd. 2009
Received: 18 March 2009
Accepted: 5 June 2009
Published: 5 June 2009
The growth-rate of an organism is an important phenotypic trait, directly affecting its ability to survive in a given environment. Here we present the first large scale computational study of the association between ecological strategies and growth rate across 113 bacterial species, occupying a variety of metabolic habitats. Genomic data are used to reconstruct the species' metabolic networks and habitable metabolic environments. These reconstructions are then used to investigate the typical ecological strategies taken by organisms in terms of two basic species-specific measures: metabolic variability - the ability of a species to survive in a variety of different environments; and co-habitation score vector - the distribution of other species that co-inhabit each environment.
We find that growth rate is significantly correlated with metabolic variability and the level of co-habitation (that is, competition) encountered by an organism. Most bacterial organisms adopt one of two main ecological strategies: a specialized niche with little co-habitation, associated with a typically slow rate of growth; or ecological diversity with intense co-habitation, associated with a typically fast rate of growth.
The pattern observed suggests a universal principle where metabolic flexibility is associated with a need to grow fast, possibly in the face of competition. This new ability to produce a quantitative description of the growth rate-metabolism-community relationship lays a computational foundation for the study of a variety of aspects of the communal metabolic life.
Variations in growth rate are observed both within and between species, reflecting, respectively, regulatory-level and genomic-level adaptations [1–4]. Since the rate of bacterial growth is determined by metabolic factors such as the rate and yield of ATP production , variations in growth rate are bound to be associated with metabolic capabilities and constraints. Several examples have demonstrated, at the single species level, that growth rate is affected by the availability of environmental resources and the level of competition in a given environment [5–8]. Comparative-growth studies have pointed to several metabolic and regulatory genes that are under selective pressure for accelerated growth - for example, genes involved in the transport of essential substrates in highly-competitive Escherichia coli populations . However, such comparative growth studies are typically restricted to species that occupy similar ecological niches, potentially missing the impact of genomic adaptations that may vary across different niches and lifestyles. To this day, the genome design principles underlying the association between growth rate and metabolic adaptations have not yet been established at a global, cross-species scale.
A comprehensive cross-species analysis, beyond a comparative study of organisms sharing a similar ecological niche, of genomic traits that are associated with the potential growth rates of bacterial organisms was made possible due to a recent list of minimal generation times of a wide spectrum of bacterial species [10, 11]. Previously, these doubling-time data have led to the important finding that variations between genes involved in translation and transcription influence growth rate [10, 11]. Here we focus on the influence of genomic-derived metabolic properties. We use genomic information to generate second-order (network-based) metabolic knowledge through the reconstruction of metabolic networks, and third-order environmental knowledge through the reconstruction of habitable metabolic environments for the species studied . Then, species-specific environmental information is further exploited to estimate the level of competition encountered by each organism according to the potential ability of other species to thrive in similar habitats . Through converting genomic data to environmental and communal information, this study examines factors that potentially underlie growth rates across all these levels, through the analysis of the metabolic networks and environments of 528 contemporary sequenced bacterial species, where growth rate data were available for 113 of these species (Additional data file 1).
Results and discussion
Growth rate is associated with basic genomic and environmental attributes
Correlation (P-value) versus duplication time
Significance of difference between slow and fast growers†
Total (N = 113)
*Non-obligatory symbiont only (N = 77)
Total (N = 113)
§Non-obligatory symbionts (N = 77)
Genome size (bp)
(S: 2,695,676, F: 3,402,099)
(S: 3,614,838, F: 3,479,053)
(S: 326, F: 410)
(S: 408, F: 431)
regulatory genes 
(S: 0.03, F: 0.05)
(S: 0.04, F: 0.05)
environmental complexity 
(S: 3, F: 4)
(S: 0.006, F: 0.02)
(S: 0.008, F: 0.02)
(S: 0.007, F: 0.03)
(S: 0.01, F: 0.004)
(S: 14, F: 27)
(S: 20, F: 31)
(S: 39, F: 72)
(S: 50, F: 85)
Typical duplication time of bacterial organisms according to their mode of respiration
Number of species*
Mean duplication time
Median duplication time
Mean network size
Modeling metabolic-environmental attributes
Growth rate is associated with the level of metabolic variability
The 'environmental scope index' (ESI) of a species is defined as the fraction of environments in which it is viable. The ESI measure of metabolic variability is positively correlated with both genome size (0.4, P = 3e-6, Spearman) and metabolic network size (0.6, P = 1e-10, Spearman). It is also positively correlated with the data-driven general environmental-diversity measures examined above (fraction of regulatory genes, 0.32, P = 0.008; estimate of environmental complexity, 0.23, P = 0.01; Spearman), hence providing support for the ecological plausibility of the model. There is a significant negative correlation between the ESI and doubling time in the complete dataset (-0.25) and, notably, the differences between the ESI scores of slow and fast growing bacteria remain significant also after excluding obligatory symbionts (Table 1). Thus, there is a general association between broader metabolic capacities and faster maximal growth rates, extending the initial observations concerning the fast growth rate of facultative bacteria and implying that the metabolic versatility of species is better associated with their growth rate than other, more general environmental characteristics. This result, as all other reported correlations, remains valid when using the alternative collection of random environments described above (Table 1; Materials and methods). The negative correlation between ESI and duplication time is also maintained in species that are evenly distributed among different habitat types and taxonomic groups (Additional data file 3). Pseudomonas aeruginosa, an organism with a high ESI score (Additional data file 1) provides an example of fast growth rate in a generalist, possessing broad metabolic capabilities that allow it to successfully grow in diverse environments . However, the association between fast growth and metabolic flexibility is not at all obvious, as one may assume that living in a specific niche habitat would enable an organism to specialize and adapt towards a fast growth solution. Indeed, Desulfotalea psychrophila, an organism with a low ESI score (Additional data file 1), provides such an example. It is a sulfate-reducing extremophilic bacterium, thriving in extreme conditions (cold arctic sediments), and exhibiting metabolic and environmental specialization .
Growth rate is associated with the level of co-habitation
If metabolic specialization does not preclude fast growth, how then can we explain the slow growth of most specialists? An emerging hypothesis is that such organisms face weak competition. Conversely, organisms that occupy a large variety of metabolic environments face a larger number of co-inhabiting species, which in turn may exert selection pressure for maintaining higher growth rates. To test this hypothesis, we used the 'co-habitation score' (CHS) vector (deduced from the environmental viability matrix), denoting the number of species that co-populate each viable environment of a given species. This vector can serve as an indication of the level of competition encountered by a species in its habitats. We focus on each species' most populated niche (maximal-CHS) and most sparsely populated one (minimal-CHS). The minimal-CHS is not significantly correlated with either ESI or doubling time. In contrast, the maximal-CHS exhibits a significant inverse relationship with duplication time (Table 1) - that is, faster growth rates are observed in richly populated, competitive environments. The maximal-CHS also displays a highly marked positive correlation with metabolic variability (P-value < 1e-3, computed by comparing to random; Additional data file 3). That is, a species' metabolic flexibility tends to erode when it populates only sparsely populated, non-competitive environments. This result remains valid when using an alternative collection of random environments (Table 1; Materials and methods). The negative correlation between maximal-CHS and duplication time is also maintained in species that are evenly distributed among different habitat types and taxonomic groups (Additional data file 3). The relevance of maximal- and minimal-CHS to growth rate can be put in a biological context by considering the lifestyle of the pathogen Staphylococcus aureus: inside a host-cell (where no competition with other bacterial species is encountered) it exhibits a far slower growth rate than in the more competitive environment of the human skin [22, 23].
Delineating major ecological strategies
Beyond these specific examples, the characterization of growth strategies based on the intricate interplay between the ESI-maximal-CHS values suggests that when taken together, these values can be used for predicting growth strategies. Using ESI and maximal-CHS values retrieved from the 113 species that were included in the original analysis, we trained a support vector machine (SVM) classifier that assigns bacterial species into one of two extreme growth classes (either fast or slow; see Materials and methods). We tested the generalization ability of the classifier in a cross-validation setting, obtaining an average receiver operating characteristic (ROC) score of 0.75 (Materials and methods). We then obtained growth rate data for Parachlamydia UWE25 , an endosymbiont of amoeba from the chlamydiae group, and for Bacillus thuringiensis, a widely distributed bacteria ; for both species, growth rate data have not so far been included in our analysis. Parachlamydia UWE25 is an obligate intracellular bacterium that exhibits reduced central metabolic and biosynthetic pathways, and is auxotrophic for most amino acids and nucleotides . In accordance with the lifestyle of this organism, we compute for this species a low environmental diversity and low competition scores (ESI, 0.002; maximal-CHS, 0; Additional data file 1). For the ubiquitous B. thuringiensis we compute high environmental diversity and high competition scores (ESI, 0.09; maximal-CHS, 59; Additional data file 1). We applied the SVM classifier to characterize the growth rates of these species; in accordance with experimental data, B. thuringiensis falls into the fast-growing category and Parachlamydia UWE25 falls into the slow-growing category (corresponding to the experimentally observed doubling times of 40 minutes  and 48 hours, respectively ).
This paper presents the first large-scale rigorous computational exploration of the ecological strategies of a species in association with its growth rate, lifestyle and metabolic capabilities. Several limitations of this analysis should be mentioned: first, the estimation of the growth environments and their viability is based on a topological network-based computation, which is obviously a first-approximation model of the underlying biology. Second, co-habitation involves additional facets of interspecies interactions beyond competition, notably cooperation and symbiosis. Bearing this in mind, the metabolic environmental model correctly captures several patterns already observed in data-driven biological habitats, testifying to its ecological plausibility.
We find that growth rate is significantly positively correlated with both the span of the environments and the level of competition that bacterial species encounter in their metabolic habitats. The model points to two main ecological strategies, suggesting a universal principle where metabolic flexibility is associated with a need to grow fast, possibly in the face of competition. The new ability to produce a quantitative description of the growth rate-metabolism-community relationship lays a computational foundation for the study of a variety of aspects of communal metabolic life. With the growing recognition that bacteria should be better studied in the context of their ecological niche and communities, future computational approaches should take into account the complex interrelationships between organisms. Such approaches are likely to become increasingly helpful for studying various aspects of microbial life within naturally occurring ecological habitats.
Materials and methods
Metabolic data were collected from the Kyoto Encyclopedia of Genes and Genomes (KEGG)  (release 46) for 528 bacterial organisms. Out of these species, information describing the maximal doubling time was available for 113 species (downloaded from ). We constructed the metabolic networks of bacterial organisms following the approach outlined in . To download networks, see Additional data file 3.
Construction of metabolic environments
Metabolic growth environments were inferred using the seed algorithm developed by . This algorithm predicts the set of exogenously acquired compounds, given the metabolic network. We additionally examined alternative approaches for generating other biologically plausible sets of random metabolic environments. One such alternative approach for generating random environments is to construct 528 shuffled seed environments - that is, maintaining an approximation of the original metabolite representation over all seeds. Each metabolite in the original seed environments is randomly assigned to the shuffled environments where its representation over all environments is 1.05 times that of its original representation. That is, if a certain metabolite has, for example, 20 appearances over all seeds, then it is randomly assigned to 21 out of 528 environments. This process is repeated for each seed metabolite. The 1.05 ratio of appearances between original and shuffled environments was chosen as it allows a similar level of habitation of the environmental viability matrix, described below (Additional data file 3).
Characterizing bacterial environments
Fraction of regulatory gene values were taken from , describing the fraction of transcription factors out of the total number of genes in the organism, an indicator of environmental variability . Environmental complexity estimated values were obtained from , where the natural environments of 117 bacterial species were categorized based on the NCBI classification for bacterial lifestyle  and ranked according to the complexity of each category (1, obligatory symbyonts; 2, specialized; 3, aquatic; 4, facultative host-associated; 5, multiple; 6, terrestrial ). Annotations for environmental complexity were available for only 68 of the 113 species for which doubling time was available. To validate the reliability of these annotations, we manually searched the literature. In two cases we changed the original annotation (from multiple to terrestrial; Additional data file 4). In addition, we searched the literature for annotations for the remaining 45 species (Additional data file 4). Together with retrieving the annotations as described above, classification of species into habitats was also done by looking for the presence of species from the dataset in environmental samples. Occurrence of species from our dataset in environmental samples was inferred according to the results of a BLAST search  of 16S RNA sequences from the 528 species in the analysis against env_nt, a comprehensive collection of sequences from environmental samples (downloaded in February 2009). We find that in the large majority of cases (30 of 33), experimental findings support the literature-based annotations (Additional data file 3).
All parameters retrieved/computed for the species in the analysis are provided in Additional data file 1.
Computing the environmental viability matrix, environmental scope index and co-habitation score
As a measure for species viability we constructed a list of 65 compounds termed 'target metabolites' (Additional data file 5), which are most likely essential for growth in most species [36–38]. A species-specific target metabolite list is formed by the intersection between the target metabolites and the metabolites that each species produces. We then tested the viability of each species over the set of 528 metabolic growth environments. Given a specific organism and an environment, an organism is considered viable in this environment if all its essential target metabolites are produced - this is examined by using a network expansion algorithm  that outputs an activated metabolic subnetwork, and verifying that the expanded subnetwork produces all target metabolites. This process yields the environmental viability matrix, whose rows denote the species, columns the metabolic environments, and binary entries the corresponding viability. From this matrix, the scope (ESI) and CHS for each species are deduced: The ESI of a species is defined as its fraction of viable environments. The CHS vector of a species records how many viable organisms populate each of its viable environments ESI and CHS values computed for all 528 species are provided (Additional data file 6).
All software used for the analysis will be provided upon request from the authors.
Constructing a support vector machine classifier
We partitioned the organisms according to their doubling time: fast and slow growers are those whose duplication time is shorter and longer by at least one standard deviation from the mean (0.48 and 9 hours, respectively; Figure S6 in Additional data file 3). Species with intermediate values were excluded from the analysis. For the remaining species, ESI and maximal-CHS values were used for training a SVM classifier with a linear kernel . We estimate the accuracy of the classifier using a ten-fold cross-validation. In this procedure, the organisms are randomly partitioned into ten distinct sets; then the class labels (slow or fast) in each set are predicted by a classifier trained on the rest of the sets. We repeated this procedure 50 times, and report the mean and standard deviation of the ROC curve . Our quality metric is the area under this curve (the ROC score).
Additional data files
The following additional data are available with the online version of this paper: a table listing genomic and ecological attributes for the 113 species in the analysis (Additional data file 1); a table listing the NCBI annotations and description of environmental samples for species that can be identified in an environmental sample (Additional data file 2); supplementary notes and figures, detailed description of all tables in the Additional data files and a table (Table S5) (Additional data file 3); a table listing the original and manually curated values of environmental complexity (Additional data file 4); a table listing the biomass target metabolites (Additional data file 5); a table listing genomic and ecological attributes for the 528 species in the metabolic analysis (Additional data file 6).
environmental scope index
receiver operating characteristic
support vector machine.
We thank Eduardo Rocha for kindly providing information on bacterial growth rates, Eyal Privman for assisting with searches against environmental databases, and Tomer Shlomi and Martin Kupiec for reading the manuscript and providing helpful feedback. This work was supported by grants from the Israeli Science Foundation (ISF), the German-Israel Foundation (GIF) and Tauber Fund to ER. SF was supported by a Long-Term EMBO Fellowship and is a fellow of the Edmond J Safra Program in Tel-Aviv University. EB is supported by the Morrison Institute for Population and Resource Studies, a grant to the Santa Fe Institute from the James S McDonnell Foundation 21st Century Collaborative Award Studying Complex Systems and by NIH Grant GM28016. RS was supported by grants from the German-Israel Foundation and ERA-NET PathoGenoMics. NY was supported by the Tel-Aviv University rector and president scholarship. UG was supported by the Bi-national Science Foundation and the German-Israeli Foundation for Research and Development.
- Berg OG, Kurland CG: Growth rate-optimised tRNA abundance and codon usage. J Mol Biol. 1997, 270: 544-550. 10.1006/jmbi.1997.1142.PubMedView ArticleGoogle Scholar
- Dong H, Nilsson L, Kurland CG: Co-variation of tRNA abundance and codon usage in Escherichia coli at different growth rates. J Mol Biol. 1996, 260: 649-663. 10.1006/jmbi.1996.0428.PubMedView ArticleGoogle Scholar
- Klappenbach JA, Dunbar JM, Schmidt TM: rRNA operon copy number reflects ecological strategies of bacteria. Appl Environ Microbiol. 2000, 66: 1328-1333. 10.1128/AEM.66.4.1328-1333.2000.PubMedPubMed CentralView ArticleGoogle Scholar
- Mager WH, Planta RJ: Coordinate expression of ribosomal protein genes in yeast as a function of cellular growth rate. Mol Cell Biochem. 1991, 104: 181-187. 10.1007/BF00229818.PubMedView ArticleGoogle Scholar
- Novak M, Pfeiffer T, Lenski RE, Sauer U, Bonhoeffer S: Experimental tests for an evolutionary trade-off between growth rate and yield in E. coli. Am Nat. 2006, 168: 242-251. 10.1086/506527.PubMedView ArticleGoogle Scholar
- Hansen SR, Hubbell SP: Single-nutrient microbial competition: qualitative agreement between experimental and theoretically forecast outcomes. Science. 1980, 207: 1491-1493. 10.1126/science.6767274.PubMedView ArticleGoogle Scholar
- Helling RB: Speed versus efficiency in microbial growth and the role of parallel pathways. J Bacteriol. 2002, 184: 1041-1045. 10.1128/jb.184.4.1041-1045.2002.PubMedPubMed CentralView ArticleGoogle Scholar
- Pfeiffer T, Schuster S, Bonhoeffer S: Cooperation and competition in the evolution of ATP-producing pathways. Science. 2001, 292: 504-507. 10.1126/science.1058079.PubMedView ArticleGoogle Scholar
- Lenski RE, Mongold JA, Sniegowski PD, Travisano M, Vasi F, Gerrish PJ, Schmidt TM: Evolution of competitive fitness in experimental populations of E. coli: what makes one genotype a better competitor than another?. Antonie Van Leeuwenhoek. 1998, 73: 35-47. 10.1023/A:1000675521611.PubMedView ArticleGoogle Scholar
- Rocha EP: Codon usage bias from tRNA's point of view: redundancy, specialization, and efficient decoding for translation optimization. Genome Res. 2004, 14: 2279-2286. 10.1101/gr.2896904.PubMedPubMed CentralView ArticleGoogle Scholar
- Couturier E, Rocha EP: Replication-associated gene dosage effects shape the genomes of fast-growing bacteria but only for transcription and translation genes. Mol Microbiol. 2006, 59: 1506-1518. 10.1111/j.1365-2958.2006.05046.x.PubMedView ArticleGoogle Scholar
- Borenstein E, Kupiec M, Feldman MW, Ruppin E: Large-scale reconstruction and phylogenetic analysis of metabolic environments. Proc Natl Acad Sci USA. 2008, 105: 14482-14487. 10.1073/pnas.0806162105.PubMedPubMed CentralView ArticleGoogle Scholar
- Janga SC, Babu MM: Network-based approaches for linking metabolism with environment. Genome Biol. 2008, 9: 239-PubMedPubMed CentralView ArticleGoogle Scholar
- Joseph B, Goebel W: Life of Listeria monocytogenes in the host cells' cytosol. Microbes Infect. 2007, 9: 1188-1195. 10.1016/j.micinf.2007.05.006.PubMedView ArticleGoogle Scholar
- Mira A, Ochman H, Moran NA: Deletional bias and the evolution of bacterial genomes. Trends Genet. 2001, 17: 589-596. 10.1016/S0168-9525(01)02447-7.PubMedView ArticleGoogle Scholar
- Backhed F, Ley RE, Sonnenburg JL, Peterson DA, Gordon JI: Host-bacterial mutualism in the human intestine. Science. 2005, 307: 1915-1920. 10.1126/science.1104816.PubMedView ArticleGoogle Scholar
- Tringe SG, von Mering C, Kobayashi A, Salamov AA, Chen K, Chang HW, Podar M, Short JM, Mathur EJ, Detter JC, Bork P, Hugenholtz P, Rubin EM: Comparative metagenomics of microbial communities. Science. 2005, 308: 554-557. 10.1126/science.1107851.PubMedView ArticleGoogle Scholar
- Mavromatis K, Ivanova N, Barry K, Shapiro H, Goltsman E, McHardy AC, Rigoutsos I, Salamov A, Korzeniewski F, Land M, Lapidus A, Grigoriev I, Richardson P, Hugenholtz P, Kyrpides NC: Use of simulated data sets to evaluate the fidelity of metagenomic processing methods. Nat Methods. 2007, 4: 495-500. 10.1038/nmeth1043.PubMedView ArticleGoogle Scholar
- Parter M, Kashtan N, Alon U: Environmental variability and modularity of bacterial metabolic networks. BMC Evol Biol. 2007, 7: 169-10.1186/1471-2148-7-169.PubMedPubMed CentralView ArticleGoogle Scholar
- Stover CK, Pham XQ, Erwin AL, Mizoguchi SD, Warrener P, Hickey MJ, Brinkman FS, Hufnagle WO, Kowalik DJ, Lagrou M, Garber RL, Goltry L, Tolentino E, Westbrock-Wadman S, Yuan Y, Brody LL, Coulter SN, Folger KR, Kas A, Larbig K, Lim R, Smith K, Spencer D, Wong GK, Wu Z, Paulsen IT, Reizer J, Saier MH, Hancock RE, Lory S, et al: Complete genome sequence of Pseudomonas aeruginosa PA01, an opportunistic pathogen. Nature. 2000, 406: 959-964. 10.1038/35023079.PubMedView ArticleGoogle Scholar
- Rabus R, Ruepp A, Frickey T, Rattei T, Fartmann B, Stark M, Bauer M, Zibat A, Lombardot T, Becker I, Amann J, Gellner K, Teeling H, Leuschner WD, Glöckner FO, Lupas AN, Amann R, Klenk HP: The genome of Desulfotalea psychrophila, a sulfate-reducing bacterium from permanently cold Arctic sediments. Environ Microbiol. 2004, 6: 887-902. 10.1111/j.1462-2920.2004.00665.x.PubMedView ArticleGoogle Scholar
- Vesga O, Groeschel MC, Otten MF, Brar DW, Vann JM, Proctor RA: Staphylococcus aureus small colony variants are induced by the endothelial cell intracellular milieu. J Infect Dis. 1996, 173: 739-742.PubMedView ArticleGoogle Scholar
- von Eiff C: Staphylococcus aureus small colony variants: a challenge to microbiologists and clinicians. Int J Antimicrob Agents. 2008, 31: 507-510. 10.1016/j.ijantimicag.2007.10.026.PubMedView ArticleGoogle Scholar
- Gomez-Valero L, Rocha EP, Latorre A, Silva FJ: Reconstructing the ancestor of Mycobacterium leprae: the dynamics of gene loss and genome reduction. Genome Res. 2007, 17: 1178-1185. 10.1101/gr.6360207.PubMedPubMed CentralView ArticleGoogle Scholar
- Cole ST, Eiglmeier K, Parkhill J, James KD, Thomson NR, Wheeler PR, Honore N, Garnier T, Churcher C, Harris D, Mungall K, Basham D, Brown D, Chillingworth T, Connor R, Davies RM, Devlin K, Duthoy S, Feltwell T, Fraser A, Hamlin N, Holroyd S, Hornsby T, Jagels K, Lacroix C, Maclean J, Moule S, Murphy L, Oliver K, Quail MA, et al: Massive gene decay in the leprosy bacillus. Nature. 2001, 409: 1007-1011. 10.1038/35059006.PubMedView ArticleGoogle Scholar
- Kapatral V, Anderson I, Ivanova N, Reznik G, Los T, Lykidis A, Bhattacharyya A, Bartman A, Gardner W, Grechkin G, Grechkin G, Zhu L, Vasieva O, Chu L, Kogan Y, Chaga O, Goltsman E, Bernal A, Larsen N, D'Souza M, Walunas T, Pusch G, Haselkorn R, Fonstein M, Kyrpides N, Overbeek R: Genome sequence and analysis of the oral bacterium Fusobacterium nucleatum strain ATCC 25586. J Bacteriol. 2002, 184: 2005-2018. 10.1128/JB.184.7.2005-2018.2002.PubMedPubMed CentralView ArticleGoogle Scholar
- Molofsky AB, Swanson MS: Differentiate to thrive: lessons from the Legionella pneumophila life cycle. Mol Microbiol. 2004, 53: 29-40. 10.1111/j.1365-2958.2004.04129.x.PubMedView ArticleGoogle Scholar
- Broderick NA, Raffa KF, Handelsman J: Midgut bacteria required for Bacillus thuringiensis insecticidal activity. Proc Natl Acad Sci USA. 2006, 103: 15196-15199. 10.1073/pnas.0604865103.PubMedPubMed CentralView ArticleGoogle Scholar
- Horn M, Collingro A, Schmitz-Esser S, Beier CL, Purkhold U, Fartmann B, Brandt P, Nyakatura GJ, Droege M, Frishman D, Rattei T, Mewes HW, Wagner M: Illuminating the evolutionary history of chlamydiae. Science. 2004, 304: 728-730. 10.1126/science.1096330.PubMedView ArticleGoogle Scholar
- Donovan WP, Tan Y, Slaney AC: Cloning of the nprA gene for neutral protease A of Bacillus thuringiensis and effect of in vivo deletion of nprA on insecticidal crystal protein. Appl Environ Microbiol. 1997, 63: 2311-2317.PubMedPubMed CentralGoogle Scholar
- Kanehisa M, Goto S: KEGG: Kyoto Encyclopedia of Genes and Genomes. Nucleic Acids Res. 2000, 28: 27-30. 10.1093/nar/28.1.27.PubMedPubMed CentralView ArticleGoogle Scholar
- Cano DA, Pucciarelli MG, Martinez-Moya M, Casadesus J, Garcia-del Portillo F: Selection of small-colony variants of Salmonella enterica serovar typhimurium in nonphagocytic eucaryotic cells. Infect Immun. 2003, 71: 3690-3698. 10.1128/IAI.71.7.3690-3698.2003.PubMedPubMed CentralView ArticleGoogle Scholar
- Madan Babu M, Teichmann SA, Aravind L: Evolutionary dynamics of prokaryotic transcriptional regulatory networks. J Mol Biol. 2006, 358: 614-633. 10.1016/j.jmb.2006.02.019.PubMedView ArticleGoogle Scholar
- Entrez Genome Project. [http://www.ncbi.nlm.nih.gov/genomes/lproks.cgi]
- Altschul SF, Madden TL, Schaffer AA, Zhang J, Zhang Z, Miller W, Lipman DJ: Gapped BLAST and PSI-BLAST: a new generation of protein database search programs. Nucleic Acids Res. 1997, 25: 3389-3402. 10.1093/nar/25.17.3389.PubMedPubMed CentralView ArticleGoogle Scholar
- Feist AM, Henry CS, Reed JL, Krummenacker M, Joyce AR, Karp PD, Broadbelt LJ, Hatzimanikatis V, Palsson BO: A genome-scale metabolic reconstruction for Escherichia coli K-12 MG1655 that accounts for 1260 ORFs and thermodynamic information. Mol Syst Biol. 2007, 3: 121-10.1038/msb4100155.PubMedPubMed CentralView ArticleGoogle Scholar
- Becker SA, Palsson BO: Genome-scale reconstruction of the metabolic network in Staphylococcus aureus N315: an initial draft to the two-dimensional annotation. BMC Microbiol. 2005, 5: 8-10.1186/1471-2180-5-8.PubMedPubMed CentralView ArticleGoogle Scholar
- Oh YK, Palsson BO, Park SM, Schilling CH, Mahadevan R: Genome-scale reconstruction of metabolic network in Bacillus subtilis based on high-throughput phenotyping and gene essentiality data. J Biol Chem. 2007, 282: 28791-28799. 10.1074/jbc.M703759200.PubMedView ArticleGoogle Scholar
- Ebenhoh O, Handorf T, Heinrich R: Structural analysis of expanding metabolic networks. Genome Inform. 2004, 15: 35-45.PubMedGoogle Scholar
- Boser BE, Guyon IM, Vapnik VN: A training algorithm for optimal margin classifiers. Proceedings of the Fifth Annual ACM Conference on Computational Learning Theory (COLT 1992): July 27-29, 1992. 1992, Pittsburgh, PA, USA: ACM Press, 144-152.Google Scholar
- Hanley JA, McNeil BJ: The meaning and use of the area under a receiver operating characteristic (ROC) curve. Radiology. 1982, 143: 29-36.PubMedView ArticleGoogle Scholar
- Metabolic Networks. [http://www.cs.tau.ac.il/~jonatha6/publications/113_dt_networks_html.zip]
This article is published under license to BioMed Central Ltd. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.