eMLSA

Databases

Assigning bacterial strains to species via the Internet – Electronic taxonomy

www.eMLSA.net provides a portal for the electronic taxonomy of bacteria, providing a common format and software for assigning strains to species via the Internet.

Electronic taxonomy contrasts with the current approach for distinguishing species within a genus, and for defining new species, which is based on polyphasic taxonomy, an approach that incorporates all available phenotypic and genotypic data into a consensus classification (Vandamme et al., 1996).

For further information and instructions please see the Instruction pages.

Software Requirements

Browser:

JAVA:

If you experience problems using the site please see the Troubleshooting guide here.

Introduction.
Primers, PCR Conditions and sequence trimming.
Entering your sequences.

Reference Tree View.

Collapsing nodes.

Locus View.
Database Query.
Downloading Data.

Submission of Strains.
System Requirements.
Acknowledgements.
References.

Introduction

eMLSA.net provides a portal for the electronic taxonomy of bacteria, providing a common format and software for assigning strains to species via the Internet.

The accepted genotypic method for defining species is based on overall genomic relatedness, such that strains which share approximately 70% of more relatedness using DNA-DNA hybridization, under standard conditions, are considered to be members of the same species (Wayne et al., 1987). In recent years there has been an interest in improving the way molecular data are used to help define species. In particular, there has been a move away from the cumbersome DNA-DNA hybridization procedure and fixed or semi-fixed cut-off values for defining species, and an increased interest in other molecular approaches to identifying and circumscribing bacterial species (Stackebrandt et al., 2002).

One newer approach is to observe the distribution of a large number of strains of closely related species in sequence space and to identify clusters of strains that are well resolved from other clusters. This approach has been developed by using the concatenated sequences of multiple core (house-keeping) genes to assess clustering patterns, and has been called multilocus sequence analysis (MLSA; Gevers et al., 2005; Hanage et al., 2006), or multilocus sequence phylogenetic analysis (Nørskov-Lauritsen et al.,2005). MLSA has been used successfully to explore clustering patterns among large numbers of strains assigned to very closely-related species by current taxonomic methods (Godoy et al., 2003; Hanage et al., 2005a,b, 2006; Hoshino et al., 2005, Bennett et al., 2007; Kilian et al., 2008; Bishop et al., submitted), to look at the relationships between small numbers of strains within a genus (Martens et al., 2008), or within a broader taxonomic grouping (Sawabe et al., 2007), and to address specific taxonomic questions (Postic et al., 2007; Thompson et al. 2007). More generally, MLSA can be used to ask whether bacterial species exist – that is, to observe whether large populations of similar strains invariably fall into well resolved clusters, or whether in some cases there is a genetic continuum in which clear separation into clusters is not observed.

eMLSA is the implementation of the MLSA approach to the assignment of strains to species clusters via the internet. eMLSA requires the generation of a large database of the sequences of the multiple house-keeping loci (typically about seven loci) from multiple strains of a set of related species of interest. It also requires software that can concatenate the sequences and produce a tree, which shows the patterns of clustering of the sequences. By comparing the species assignments of the strains within each sequence cluster, including the position of the type strain of each species, the sequence clusters can be assigned as species clusters. Having generated a database and assigned the species clusters, the sequences of the multiple loci of a query strain are submitted to the relevant eMLSA.net subsite and the position of the strain on the reference tree is returned, with the species assignment. Additional software at eMLSA.net allows the most similar concatenated sequences in the database(and the species assigned to these closest matches) to be returned and for the source of each allele to be assigned as resident to that species or possibly imported from a related species.

Microbiologists are encouraged to submit new strains, and the corresponding sequences, to the curator at eMLSA.net as this expands the database, but also has the potential to identify new sequence clusters, which may be assigned as new species. Our view is that electronic taxonomy, as implemented at eMLSA.net, provides a way of assigning strains to species that is at least as valid as other methods. At the very least it provides a hypothesis about the relationships between sequence clusters generated using core genes and species clusters that can be tested, revised or refuted by the addition of phenotypic, genotypic or ecological data.

At present, eMLSA.net databases are available for viridans group streptococci, a group where species assignment by more traditional means is problematic, for Burkholderia pseudomallei and related species, and for Borrelia species.

Please contact David Aanensen (d.aanesen[at]imperial.ac.uk) if you wish to host a MLSA subsite at eMLSA.net.

Primers, PCR Conditions and sequence trimming

Primer	Gene product	Sequence (5'-3')^*	Primer length (bp)	Trimmed fragment size (bp)	Annealing temperature (^oC)
map-up map-dn	Methionine aminopeptidase	GCWGACTCWTGTTGGGCWTATGC TTARTAAGTTCYTTCTTCDCCTTG	23 24	348	55 55
pfl-up pfl-dn	pyruvate formate lysase	AACGTTGCTTACTCTAAACAAACTGG ACTTCRTGGAAGACACGTTGWGTC	26 24	351	55 55
ppaC-up ppaC-dn	Inorganic pyrophosphatase	GACCAYAATGAATTYCARCAATC TGAGGNACMACTTGTTTSTTACG	23 23	552	50 50
pyk-up pyk-dn	Pyruvate kinase	GCGGTWGAAWTCCGTGGTG GCAAGWGCTGGGAAAGGAAT	19 20	492	50 50
rpoB-up rpoB-dn	RNA polymerase beta subunit	AARYTIGGMCCTGAAGAAAT TGIARTTTRTCATCAACCATGTG	20 22	516	50 50
sodA-up sodA-dn	Superoxide dismutase	TRCAYCATGAYAARCACCAT ARRTARTAMGCRTGYTCCCARACRTC	20 26	378	50 50
tuf-up tuf-dn	Elongation factor Tu	GTTGAAATGGAAATCCGTGACC GTTGAAGAATGGAGTGTGACG	22 21	426	55 55

^*Sequences are shown using the IUPAC codes for sites where degeneracy was introduced.

To view a single example of each sequence please click the following links - COMING SHORTLY!!!

Entering your sequences

After choosing the eMLSA database from the links on the left hand side of the page, you will be presented with a form consisting of a number of text boxes corresponding to the loci used for MLSA within the species group. The gene name can be seen above each textbox. Paste each trimmed sequences into the corresponding textbox as one continuous string (ie including no spaces.)

Error messages are returned should your sequence contain any non-DNA characters or be of incorrect length for that locus.

Once all seven sequences are entered correctly, the borders of the textboxes will turn green indicating that the sequences are correctly entered.

Should the genome sequences be available for strains of some species within a species group, a checkbox is present which, when checked, will include the concatenated sequences from these genomes when producing the reference tree.
By default genome sequences are NOT included in the reference tree.

Click ‘submit’ to proceed.

Click ‘Click to view tree’ which will take you to the Tree View page.

Navigation is undertaken using the links in the ‘Navigation’ Menu and four links are provided as follows.

Reference Tree View.
Locus View.
Database Query.
Downloading Data.

Reference Tree View

On submitting your seven sequences, they are concatenated, and the concatenate is added to the database that includes the sequences of all strains within the species group. These sequences are then aligned and a distance matrix is produced. A Neighbor joining tree based on this matrix is then viewable on the website.

This process can take time (typically about 30 seconds) and loading indicators are shown on the page while processing of the data takes place.

The results window consists of four sections as follows –

Tree

PhyloWidget displays by default a radial tree with your query indicated as ‘YOUR QUERY’ If you cannot see your query strain you should zoom-in (see below), or you can highlight your query strain by typing ‘YOUR QUERY’ into the Search Box at the top of the PhyloWidget window.

Tree Controls

The size of the terminal node, or the text size of the strain labels, can be changed using the two slider bars below the tree. Simply move the slider to the right to increase either.
Labels (Strain names) can be switched on or off using the links below the slider bars.
By default a radial tree display is returned. However, you can change the layout to produce the tree in a circular, or rectangular tree format using the links provided below the slider bars.
The small section to the left of the tree window contains three options -

	Arrow (shortcut key – A) allows you, when selected, to point and select nodes on the tree.
	Scroll (shortcut key – S) click and drag to move the tree around the Tree View window.
	Zoom (shortcut key – Z) click and drag up or down to zoom in or out of the tree.

The drop-down menus at the top of the Tree View Window allow various further actions to be undertaken on the tree, including exporting images for printing and saving the NEWICK format file. For further instructions please see – http://www.phylowidget.org

Closest matches

Once loaded, the species names assigned to the five best matches in the database to your concatenated query sequence are displayed, including the % similarities to the query strain. If you are using Firefox (See Site Requirements) hyperlinks are present on the five returned strain names, which allows any of these strains to be highlighted on the tree when clicked.

Navigation between pages is undertaken using the ‘Navigation’ Menu. As detailed above

Collapsing nodes on the tree.

Any node on the tree can be collapsed allowing easier viewing of smaller subsets of sequences. Using the pointer tool, click on the node that you wish to collapse. A context menu appears offering a number of options as in the following screenshot.

Select 'Tree Edit' and then 'Collapse' The node selected will then be collapsed and in its place a yellow triangle will appear as in the following screenshot.

You may need to zoom out to see the full tree and adjust the terminal node and text sizes using the sliders or alternatively you can select 'View' and 'Zoom to full' (shortcut - Ctrl+F) from the Phylowidget menu.

To uncollapse the node, simply click the yellow triangle.

Locus View

The Locus View contains sections which, once loaded, display the closest matches within the database to each of your pasted in allele sequences.
Each locus section contains the top 5 database matches including the assigned species, the strain name and the % identity to your sequence.
Based on the overall species assignment of your concatenated sequence, this allows the allele sequences at each locus to be assigned as ‘resident’,’ resident compatible’ or ‘foreign’ as follows.

Based on a hypothetical set of sequences where the overall species assignment (based on the concatenated sequences) assigns the strain as Streptococcus australis we can envision the following circumstances for each locus.

Resident : The sequence at the locus is assigned to the same species as that assigned to the concatenated sequence.

resident

The overall species assignment is S.australis and the query sequence for the pfl locus falls into the S.australis cluster, which is well resolved from other species on the single gene tree. You can click the button to view the individual gene tree to see where the query sequence clusters on the pfl gene tree.

Resident Compatible : The sequence at the locus is compatible with being from the same species as that assigned to the concatenated sequence.

resident

The overall species assignment is S.australis. However, even though the top match at the ppaC locus returns S.australis, the S.australis sequences on the individual gene tree are not well resolved from S.infantis. Consequently, the 2nd, 3rd and 4th matches are to S.infantis rather than S.australis. The Query ppaC sequence is assigned as ’Resident compatible’ as it is almost certainly a native S.australis allele, although the lack of resolution from S.infantis makes it a theoretical possibility that it has been imported from the latter species. Clicking the button shows the ppaC gene tree to see where the query sequence clusters.

Foreign : The allele appears to have been introduced from a related species.

resident

The overall species assignment is S.australis. However, the query sequence at the map locus is most similar to S.infantis strains. As the S.infantis sequences are well resolved from S.australis sequences on the map gene tree the map query sequence is assigned as ’foreign’, presumably having been imported from an S.infantis strain.

Database Query

The database query allows the strains submitted to the eMLSA scheme to be interrogated and further details returned. Information included will depend on the species group under investigation but a list of searchable columns can be viewed in the dropdown lists of the search form.

Enter your search terms in the boxes provided. After submitting, results appear below the search boxes.

Downloading data

You can download all data used to generate trees and related epidemiological data for all strains within the database.

All sequences (concatenated or for individual loci) can be downloaded in MEGA, FASTA, Tab-delimited or comma separated (CSV) formats, for use in other analysis programs.

Simply click the corresponding arrow to download the sequence data.

From within Phylowidget you can also download the NEWICK data that describes the tree shown. Simply click (on the Phylowidget Menu) 'File' and then 'Save Tree'.

Submission of Strains to eMLSA.net

The submission of MLSA data on strains within, or closely allied to, groups of species included in the different MLSA schemes is strongly encouraged as it builds the database and potentially can identify new species clusters.

Please email the correctly trimmed sequences for all loci, and the forward and reverse sequencer trace files, to the curator and your proposed species assignment for the strain. The curator will check the data and will request further details of the strain before entering it in the relevant MLSA database.

The names and email addresses of the curators are given on the front page of each MLSA scheme.

Requirements - JAVA, Web browser and Troubleshooting.

eMLSA.net is built using AJAX, Perl and PHP. Phylowidget is built using Processing and as such requires JAVA to be installed on your computer along with a Javascript enabled Web browser. Realistically, most modern Operating systems come with the correct version of JAVA and a web browser capable of running the site. However, if you experience any difficulties please refer to the following specifics -

JAVA- PhyloWidget requires Java 1.5 or higher (this is also known as Java 5.0). Most modern systems, including Windows XP andMac OS X should have this installed; if for some reason it appears that Java is not set up correctly on your system, simply visit http://www.Java.com to download the latest version.
Automatic JAVA detection is present on this website and the lower box on the frontpage will detect whether JAVA is installed and which version you have.

Web Browser - Below is a table containing recommended browsers for viewing eMLSA.net. We recommend using Firefox if possible - it is free to download and available for most commonly used Operating Systems.
For specific known browser issues when using Phylowidget please see here

OS	Browser	Support(and known issues)	Download
Windows	Firefox	3.0+ OK
	Internet Explorer	v5.0+ OK, some issues with JAVA refresh on IE8beta+
	Safari	3.0+ OK
Mac OSX	Firefox	3.0+ OK
Mac OSX	Safari	3.0+ OK
Linux	Firefox	3.0+ OK

We encourage you to let us know if you experience difficulties so that we can improve eMLSA.net.
Should you have any issues please email David Aanensen.

Troubleshooting

Browser crash / freeze-up
With certain versions of Firefox, Java, and Windows XP or Vista, we have observed the browser crashing entirely when PhyloWidget or any other Java applet is run. If this is happening to you, please visit Java.com and update your version of Java.
Sluggish / strange behavior
The main point is this: Most strange or sluggish behavior in PhyloWidget can be fixed by restarting your browser!!

Usually refreshing the browser will not fix problems with sluggishness. The Java runtime environment stays loaded for the entire time the browser is running, so the only way to give PhyloWidget a "fresh start" is to completely quit the browser, and then start it back up.

To reiterate: If you're having a problem with PhyloWidget suddenly acting strange, 90% of the time restarting the browser will solve it!
OutOfMemory Error / Java heap space
If you see either of these error message while using PhyloWidget, or the program appears to putter out when working with large trees or images, then the operation you attempted may be using more memory than is allocated to Java applets by default.

The simplest solution is to download the standalone version of PhyloWidget from the homepage and run the program as an application, where more memory is allocated to the JVM.

Another option is to edit your computer-specific applet settings to allocate more memory: see the relevant Java forum thread, or search Google for java applet heap space. One set of visual instructions for Windows XP can be found here.
Error pop-ups
Some users, especially on Linux systems, have reported seeing pop-up messages mentioning Null Pointer Exceptions when trying to load PhyloWidget. If you see this, please submit a bug report to David Aanensen

Acknowledgements

The server-side version of MEGA4 has been kindly provided by Dr Sudhir Kumar, Arizona State University.
Please click the image to visit the MEGA website.

eMLSA.net thanks the developers of the following Javascript libraries-

A special thanks to Greg Jordan for adapting Phylowidget for the purposes of eMLSA.net.
For further information on PhyloWidget please click here

References

Gevers D, Cohan FM, Lawrence JG, Spratt BG, Coenye T, Feil EJ, Stackebrandt E, Van de Peer Y, Vandamme P, Thompson FL, Swings J (2005).
Opinion: Re-evaluating prokaryotic species. Nat Rev Microbiol
Click here to toggle abstract

Godoy D, Randle G, Simpson AJ, Aanensen DM, Pitt TL, Kinoshita R, Spratt BG (2003)
Multilocus sequence typing and evolutionary relationships among the causative agents of melioidosis and glanders, Burkholderia pseudomallei and Burkholderia mallei. J Clin Microbiol
Click here to toggle abstract

Multilocus sequence typing and evolutionary relationships among the causative agents of melioidosis and glanders, Burkholderia pseudomallei and Burkholderia mallei.

J Clin Microbiol. 2003 May;41(5):2068-79

Authors: Godoy D, Randle G, Simpson AJ, Aanensen DM, Pitt TL, Kinoshita R, Spratt BG

A collection of 147 isolates of Burkholderia pseudomallei, B. mallei, and B. thailandensis was characterized by multilocus sequence typing (MLST). The 128 isolates of B. pseudomallei, the causative agent of melioidosis, were obtained from diverse geographic locations, from humans and animals with disease, and from the environment and were resolved into 71 sequence types. The utility of the MLST scheme for epidemiological investigations was established by analyzing isolates from captive marine mammals and birds and from humans in Hong Kong with melioidosis. MLST gave a level of resolution similar to that given by pulsed-field gel electrophoresis and identified the same three clones causing disease in animals, each of which was also associated with disease in humans. The average divergence between the alleles of B. thailandensis and B. pseudomallei was 3.2%, and there was no sharing of alleles between these species. Trees constructed from differences in the allelic profiles of the isolates and from the concatenated sequences of the seven loci showed that the B. pseudomallei isolates formed a cluster of closely related lineages that were fully resolved from the cluster of B. thailandensis isolates, confirming their separate species status. However, isolates of B. mallei, the causative agent of glanders, recovered from three continents over a 30-year period had identical allelic profiles, and the B. mallei isolates clustered within the B. pseudomallei group of isolates. Alleles at six of the seven loci in B. mallei were also present within B. pseudomallei isolates, and B. mallei is a clone of B. pseudomallei that, on population genetics grounds, should not be given separate species status.

PMID: 12734250 [PubMed - indexed for MEDLINE]

Hanage WP, Fraser C, Spratt BG (2005)
Fuzzy species among recombinogenic bacteria. BMC Biol
Click here to toggle abstract

Hanage WP, Kaijalainen T, Herva E, Saukkoriipi A, Syrjänen R, Spratt BG (2005)
Using multilocus sequence data to define the pneumococcus. J Bacteriol
Click here to toggle abstract

Hanage WP, Fraser C, Spratt BG (2006)
Sequences, sequence clusters and bacterial species. Philos Trans R Soc Lond B Biol Sci
Click here to toggle abstract

Hoshino T, Fujiwara T, Kilian M (2005)
Use of phylogenetic and phenotypic analyses to identify nonhemolytic streptococci isolated from bacteremic patients. J Clin Microbiol
Click here to toggle abstract

Use of phylogenetic and phenotypic analyses to identify nonhemolytic streptococci isolated from bacteremic patients.

J Clin Microbiol. 2005 Dec;43(12):6073-85

Authors: Hoshino T, Fujiwara T, Kilian M

The aim of this study was to evaluate molecular and phenotypic methods for the identification of nonhemolytic streptococci. A collection of 148 strains consisting of 115 clinical isolates from cases of infective endocarditis, septicemia, and meningitis and 33 reference strains, including type strains of all relevant Streptococcus species, were examined. Identification was performed by phylogenetic analysis of nucleotide sequences of four housekeeping genes, ddl, gdh, rpoB, and sodA; by PCR analysis of the glucosyltransferase (gtf) gene; and by conventional phenotypic characterization and identification using two commercial kits, Rapid ID 32 STREP and STREPTOGRAM and the associated databases. A phylogenetic tree based on concatenated sequences of the four housekeeping genes allowed unequivocal differentiation of recognized species and was used as the reference. Analysis of single gene sequences revealed deviation clustering in eight strains (5.4%) due to homologous recombination with other species. This was particularly evident in S. sanguinis and in members of the anginosus group of streptococci. The rate of correct identification of the strains by both commercial identification kits was below 50% but varied significantly between species. The most significant problems were observed with S. mitis and S. oralis and 11 Streptococcus species described since 1991. Our data indicate that identification based on multilocus sequence analysis is optimal. As a more practical alternative we recommend identification based on sodA sequences with reference to a comprehensive set of sequences that is available for downloading from our server. An analysis of the species distribution of 107 nonhemolytic streptococci from bacteremic patients showed a predominance of S. oralis and S. anginosus with various underlying infections.

PMID: 16333101 [PubMed - indexed for MEDLINE]

Kilian M, Poulsen K, Blomqvist T, Håvarstein LS, Bek-Thomsen M, Tettelin H, Sørensen UB (2008)
Evolution of Streptococcus pneumoniae and its close commensal relatives. PLoS ONE
Click here to toggle abstract

Martens M, Dawyndt P, Coopman R, Gillis M, De Vos P, Willems A (2008)
Advantages of multilocus sequence analysis for taxonomic studies: a case study using 10 housekeeping genes in the genus Ensifer (including former Sinorhizobium). Int J Syst Evol Microbiol
Click here to toggle abstract

Nørskov-Lauritsen N, Bruun B, Kilian M(2005)
Multilocus sequence phylogenetic study of the genus Haemophilus with description of Haemophilus pittmaniae sp. nov. Int J Syst Evol Microbiol
Click here to toggle abstract

Postic D, Garnier M, Baranton G (2007)
Multilocus sequence analysis of atypical Borrelia burgdorferi sensu lato isolates--description of Borrelia californiensis sp. nov., and genomospecies 1 and 2. Int J Med Microbiol
Click here to toggle abstract

Sawabe T, Kita-Tsukamoto K, Thompson FL (2007)
Inferring the evolutionary history of vibrios by means of multilocus sequence analysis. J Bacteriol
Click here to toggle abstract

Thompson FL, Gomez-Gil B, Vasconcelos AT, Sawabe T (2007)
Multilocus sequence analysis reveals that Vibrio harveyi and V. campbellii are distinct species. Appl Environ Microbiol
Click here to toggle abstract

Multilocus sequence analysis reveals that Vibrio harveyi and V. campbellii are distinct species.

Appl Environ Microbiol. 2007 Jul;73(13):4279-85

Authors: Thompson FL, Gomez-Gil B, Vasconcelos AT, Sawabe T

Identification and classification of Vibrio species have relied upon band pattern methods (e.g., amplified fragment length polymorphism) and DNA-DNA hybridization. However, data generated by these methods cannot be used to build an online electronic taxonomy. In order to overcome these limitations, we developed the first standard multilocus sequence scheme focused on the ubiquitous and pathogenic Vibrio harveyi species group (i.e., V. harveyi, V. campbellii, V. rotiferianus, and a new as yet unnamed species). We examined a collection of 104 isolates from different geographical regions and hosts using segments of seven housekeeping genes. These two species formed separated clusters on the basis of topA, pyrH, ftsZ, and mreB gene sequences. The phylogenetic picture obtained by the other three loci, i.e., gyrB, recA, and gapA, was more complex though. V. campbellii appeared nested within V. harveyi in the recA trees, whereas V. harveyi formed a tight nested cluster within V. campbellii by gapA. The gyrB gene had no taxonomic resolution and grouped the two species together. The fuzziness observed in these three genes seems not be related to recombination but to low divergence due to the accumulation of only a few substitutions. In spite of this, the concatenated sequences provided evidence that the two species form two separated clusters. These clusters did not arise by recombination but by accumulation of point mutations. V. harveyi and V. campbellii isolates can be readily identified through the open database resource developed in this study (http://www.taxvibrio.lncc.br/). We argue that the species should be defined by evolutionary criteria. Strains of the same species will share at least 95% concatenated sequence similarity using the seven loci, and, most importantly, cospecific strains will form cohesive readily recognizable phylogenetic clades.

PMID: 17483280 [PubMed - indexed for MEDLINE]

Stackebrandt E, Frederiksen W, Garrity GM, Grimont PA, Kämpfer P, Maiden MC, Nesme X, Rosselló-Mora R, Swings J, Trüper HG, Vauterin L, Ward AC, Whitman WB (2002)
Report of the ad hoc committee for the re-evaluation of the species definition in bacteriology. Int J Syst Evol Microbiol
Click here to toggle abstract

Vandamme P, Pot B, Gillis M, de Vos P, Kersters K, Swings J (1996)
Polyphasic taxonomy, a consensus approach to bacterial systematics. Microbiol Rev
Click here to toggle abstract

Polyphasic taxonomy, a consensus approach to bacterial systematics.

Microbiol Rev. 1996 Jun;60(2):407-38

Authors: Vandamme P, Pot B, Gillis M, de Vos P, Kersters K, Swings J

Over the last 25 years, a much broader range of taxonomic studies of bacteria has gradually replaced the former reliance upon morphological, physiological, and biochemical characterization. This polyphasic taxonomy takes into account all available phenotypic and genotypic data and integrates them in a consensus type of classification, framed in a general phylogeny derived from 16S rRNA sequence analysis. In some cases, the consensus classification is a compromise containing a minimum of contradictions. It is thought that the more parameters that will become available in the future, the more polyphasic classification will gain stability. In this review, the practice of polyphasic taxonomy is discussed for four groups of bacteria chosen for their relevance, complexity, or both: the genera Xanthomonas and Campylobacter, the lactic acid bacteria, and the family Comamonadaceae. An evaluation of our present insights, the conclusions derived from it, and the perspectives of polyphasic taxonomy are discussed, emphasizing the keystone role of the species. Taxonomists did not succeed in standardizing species delimitation by using percent DNA hybridization values. Together with the absence of another "gold standard" for species definition, this has an enormous repercussion on bacterial taxonomy. This problem is faced in polyphasic taxonomy, which does not depend on a theory, a hypothesis, or a set of rules, presenting a pragmatic approach to a consensus type of taxonomy, integrating all available data maximally. In the future, polyphasic taxonomy will have to cope with (i) enormous amounts of data, (ii) large numbers of strains, and (iii) data fusion (data aggregation), which will demand efficient and centralized data storage. In the future, taxonomic studies will require collaborative efforts by specialized laboratories even more than now is the case. Whether these future developments will guarantee a more stable consensus classification remains an open question.

PMID: 8801440 [PubMed - indexed for MEDLINE]