TETRA: a web-service and a stand-alone program for the analysis and comparison of tetranucleotide usage patterns in DNA sequences

At present, a majority of the microbes from natural microbial communities cannot be transferred into pure cultures, either because proper cultivation conditions have yet to be found, or due to currently unidentified fundamental obstacles [1]. For decades, this has limited our understanding of the functioning of microbes within their natural habitats, such as their metabolic roles, interactions and dependencies.

The metagenome-approach allows for the first time to circumvent the restrictions imposed by the limited culturability of environmental microorganisms [2]. Now, DNA fragments of 40 – 150 kb of uncultured microorganisms can be cloned directly from the environment. This delivers insights into the microorganisms' genetic potentials, sometimes even allowing the reconstruction of entire genomes. Prominent example include the unexpected finding of bacteriorhodopsin in marine Gammaproteobacteria [3–6], and the almost complete reconstruction of two bacterial genomes from an acid mine drainage microbial biofilm [7], respectively.

However, the metagenome-approach is not without its limitations and problems. A major constraint is that especially small genomic fragments, as are obtained from libraries that have been constructed with fosmids or cosmids as cloning vectors, often lack suitable phylogenetic marker genes. This leads to the problem that fragments belonging to the same organism cannot be reliably identified as such unless they overlap. In order to nonetheless interrelate such genomic fragments, measures such as the (G+C)-content or BLAST hits and codon usage of the fragment's coding regions are commonly used to assess whether two unlinked fragments from a metagenome library belong to the same organism. These measures, however, can produce ambiguous or even misleading results, and should be supplemented by additional tools that assess the relatedness of the genomic fragments [8].

Since numerous studies have shown that oligonucleotide frequencies within DNA sequences exhibit species-specific patterns [9–18], comparative analysis of such oligonucleotide frequencies is a promising approach to this problem. For tetranucleotides, it has even been demonstrated that their frequencies carry an innate but weak phylogenetic signal [19]. Comparative analysis of tetranucleotide usage patterns also provides a good balance between computational requirements and attainable resolution. This makes the method particularly well-suited for use as a high-throughput method that can assist in tackling the fragment identification problem in metagenomics [8].

In order to automate and facilitate such an analysis, the TETRA software suite was developed, comprising both, a web-service and a stand-alone program.

The algorithms that are used within TETRA have been described elsewhere [8]. In brief, DNA sequences are extended by their reverse-complements to compensate for different patterns of tetranucleotide over- and underrepresentation between the leading and the lagging strand. Then, the frequencies of all 256 possible tetranucleotides are counted and the corresponding expected frequencies are calculated by means of a maximal-order Markov model from the sequences' di- and trinucleotide composition. In order to evaluate the significance of the level of over- or underrepresentation for each tetranucleotide, the divergence between the observed and expected tetranucleotide frequencies is then transferred into z-scores using an approximation published by Schbath [20, 21]. Finally, all DNA sequences are compared in pairs by computing the Pearson's correlation coefficient of their z-scores. Details on the method, its applicability and its limits are given in Teeling et al. (2004) and the TETRA online manual.

The TETRA web-service [22] has been implemented as set of PERL CGI scripts. Access is free to all users. A multi-headed FASTA file with DNA sequence data can be uploaded (actual file size limit: 2 Mb) and after having entered a valid e-mail address, the calculation can be started (Figure 1). Results are sent to the respective e-mail address as a tab-delimited crosstabulation of correlation coefficients in plain text format.

The TETRA stand-alone program can be downloaded for free from the TETRA website [23]. The current release has been implemented in REALbasic (http://www.realsoftware.com REAL Software Inc., Austin, Texas) and is available for Mac OS X. Versions for Linux and Windows are also available, but differ in details regarding their implementation and features. The counting of the tetranucleotides in the current version of TETRA stand-alone program is done by ocount – a self-written C program that has been integrated into the program.

With the worldwide ongoing programs to sequence and analyze natural communities, new approaches for sequence correlation beyond G+C content, read densities and codon usage have to be developed and made available to the users. The easy to use TETRA software will facilitate this task and provide additional decision support for, e.g., fragment assignment also when complete genomes have to be assembled in environmental sequencing projects.