Abstract

Background

Recently, there has been much interest in the use of large, concatenated multiple sequence alignments ('supermatrices') for phylogenetic analysis [1,2]. Such datasets have been shown to be useful in resolving difficult phylogenetic questions with a high degree of confidence. By combining the phylogenetic signal from multiple genes, clades can be recovered that are not recovered under analysis of any of the individual genes. Additionally, genes evolving at different rates may offer resolution at different phylogenetic levels.

Large-scale phylogenetic analyses of the type described above place a heavy burden of sequence acquisition, dataset assembly and dataset storage on the researcher. Sequences corresponding to the genes of interest must be obtained from public databases and orthology assigned. Where multiple sequences are available for a given gene in a species (as is often the case with EST datasets, for example) a consensus sequence must be derived. The sequences for each gene must then be aligned before being added to a concatenated alignment file, which may also contain commands necessary to partition the data. The forest of trees resulting from subsequent phylogenetic analysis must be associated with the relevant datasets, analysis parameters and confidence metrics. With these tasks in mind we have developed an integrated solution, TaxMan. TaxMan reduces this curatorial burden by automatically assembling and storing large aligned sequence datasets, and also storing metadata and trees resulting from phylogenetic analysis. Because of the high level of automation offered by TaxMan, datasets can be rebuilt rapidly to include new sequence data.

Implementation

TaxMan is written in Perl [see Additional file 1] and makes extensive use of modules from the BioPerl project [3]. PostgreSQL, a relational database management system (RDBMS), is used to store data. TaxMan makes use of many freely available bioinformatics tools; see the user guide [see Additional file 2] for a complete list and details of how to obtain them. The steps required to build a database of aligned genes are threefold; sequence acquisition, consensus derivation and alignment. Figure ​Figure11 gives an overview of the TaxMan workflow.

Open in a separate window

Figure 1

Diagram of the TaxMan workflow showing the steps involved in multigene phylogenetic analysis. The upper dotted box shows the stages that are carried only once, to build the database – sequence acquisition, consensus building and alignment. The lower dotted box shows the stages that are typically carried out multiple times – definition of a subset of genes and taxa, production of an alignment file, subjecting the alignment to phylogenetic analysis and storing the resulting tree.

Results

TaxMan was used to assemble a dataset for seventeen loci, including mitochondrial protein-coding (all 13 animal mitochondrial protein-coding genes) and ribosomal RNA genes (12S and 16S), and nuclear ribosomal RNA genes (SSU, LSU), from the subphylum Chelicerata. Approximately 12,000 annotated CoreNucleotide records and 74,000 EST records were downloaded from the NCBI GenBank database using Entrez [5](15/10/2005). TaxMan extracted 5,410 annotated sequences and 8,807 screened sequences using a chelicerate mitochondrial genome BLAST database. 1,204 species were included, representing 3 classes, 17 orders and 142 families. 2,398 consensus sequences were generated (1,992 from annotated sequences, 136 from EST sequences, 270 from fully sequenced mitochondrial genomes) and aligned. At this point, subsets of genes and taxa were specified and concatenated alignment files generated for phylogenetic analysis. For example, a slice was automatically specified that contained all genes and 3 representative species per order, selecting taxa with the most sequence available. The alignment file resulting from this slice was analysed to investigate the relationships between the chelicerate orders [12]. The process took approximately two hours (not including phylogenetic analysis) on a Linux server (Intel Xeon 3.4 GHz processor and 2 GB of RAM).

Discussion

We envisage the target users of TaxMan as researchers with experience of phylogenetics who are interested in assembling and curating a large set of aligned sequences and resulting trees for a taxonomic group for multiple-gene phylogeny. Using TaxMan, the user can assemble a large dataset of aligned DNA and protein sequences for multiple genes much more quickly and more completely than could be achieved manually. It can cope with very large input datasets and also facilitates selection of taxa for analysis and curation of phylogenetic results. Other phylogenetics software packages do not fulfil the role of TaxMan. Some packages carry out a single step in the phylogenetic process (for example, HAL [13] attempts to define orthologues in a sequence dataset using sequence similarity, and the Phylota [14] project contains applications for database clustering and subset definition) but lack integration. ARB [15] is a computing environment for phylogenetic analysis of DNA sequences, originally developed for RNA genes, but it cannot cope with multiple genes or construct consensus sequences. TreeBlaster [16] gathers sequence data based on BLAST similarity and uses phylogenetic software tools to construct a tree, but cannot use multiple genes, select sets of sequences for analysis, build consensus sequences or store phylogenetic results. The Arthropodan Mitochondrial Genomes Accessible database (AmiGA [17]) produces concatenated alignments for multiple genes, but requires that slices are specified prior to alignment, resulting in much greater processing times.

TaxMan is limited by design in two main ways; firstly, it is incapable of storing multiple reference taxonomies, instead taking the NCBI taxonomy as a base. Secondly, it assumes that a single consensus sequence is to be generated for each locus in each species, making it unsuitable for sampling from gene families, or from populations from within species (as found, for example, in DNA barcoding studies). Planned future directions for TaxMan include support for multiple taxonomies and for multiple sequences per locus. A mechanism allowing users to easily exclude sequences that have been incorrectly identified would be of benefit – currently, sequences must be deleted manually from the database. Routines for automatic orthology assignment, a significant problem in its own right, could also be included. The ability of TaxMan to store multiple phylogenetic trees for a given set of terminal taxa opens up additional possibilities for automated tree comparison tools.

Conclusion

There are many scenarios in which researchers might want to assemble a dataset of aligned DNA or protein sequences with a view to phylogenetic analysis. Current approaches require assembling such datasets manually or, at best, using software tools that lack integration. TaxMan greatly facilitates the various steps required for a multigene phylogenetic study by extracting sequences from public databases, building and aligning consensus sequences, choosing sets of data for analysis and storing the results of analysis. By using a relational database to store data, along with existing bioinformatics tools and a Perl framework, TaxMan allows large datasets to be assembled extremely rapidly and with full data provenance, allowing the user to concentrate on the analysis.

Availability and requirements

Project name: TaxMan

Project home page: http://www.nematodes.org/bioinformatics/Taxman/

Operating system: Linux

Programming language: Perl (5.8.6)

Other requirements: BioPerl (1.4), CPAN, PostgreSQL (8.0.8), BLAST (2.2.13), EMBOSS (3.0.0), phrap (0.990329), POA

License: GNU GPL

Restrictions: None

Authors' contributions

MJ wrote and documented the software and wrote the manuscript. MB assisted with design and testing of the software and writing of the manuscript, and supervised the project. Both authors read and approved the manuscript.

Supplementary Material

Acknowledgements

References