Species identification is essential for large-scale biodiversity monitoring and conservation [1]. Several molecular methods have been employed for biodiversity studies, but traditional methods such as allozyme analysis are usually labor-intensive and irreproducible. Because of advances in DNA-based technologies, approaches such as DNA arrays and DNA barcoding have recently gained attention. Both of these methods are based on comparative DNA sequence analysis, but they have significant differences.

Micro- and macro-arrays rely on the hybridization of short (i.e. 25 base) specific nucleotide probes to DNA from the target organism and subsequent detection of the hybridization signal. Although array-based technologies have been widely used in gene expression studies, their use in biodiversity research has been less rigorous, mainly targeting pathogenic microorganisms [2] and arrays of environmental samples [3]. Pfunder et al [4], however, have advocated an array-based method for the identification of voles and shrews for biodiversity monitoring. Although this study focuses on a limited number of species, the authors have predicted that such an approach can be used for the development of a so called 'Mammalia Chip', in the case of mammalian species, or even a 'Biodiversity Chip' for monitoring key species of different taxa from bacteria to mammals [4].

Species identification by DNA barcoding is based on sequencing a short standardized genomic region of the target specimen and comparing this information to a sequence library from known species [5]. The proposed standard barcode sequence for animal species is a 650-bp fragment of the mitochondrial gene cytochrome c oxidase I (COI, cox1). This DNA barcode has successfully been used for the identification of species in various vertebrate and invertebrate groups from birds to Lepidoptera [6–8], and in different geographical settings from the arctic to the tropics [6, 9]. Additionally, smaller fragments (i.e. 100 bases) of the standard COI barcode – 'mini-barcodes' – have been shown to be effective for species identification in specimens whose DNA is degraded or potentially in other situations where obtaining a full-length barcode is not feasible [10]. Barcoding is now being extended to other groups such as fungi, plants and protists, and the Barcode of Life Initiative has gained international momentum by the establishment of the Consortium for the Barcode of Life (CBOL), which plans to assemble DNA barcode libraries for all fish and birds [11].

Here, we compare the design and applicability of both array-based and barcoding platforms for specimen identification in mammalian species. We have chosen mammals because they constitute an important target for biodiversity studies and include many endangered species. However, mammalian species have not been broadly targeted for developing array-based or barcoding identification systems previously. A rapid identification method will aid in the tracking of illegal trafficking of mammalian species and their tissues. We have selected two mitochondrial loci for our analysis: COI – the proposed standard animal DNA barcode – and cytochrome b (cytb), which is commonly used as a species-level marker and particularly so in mammalian biosystematics [4, 12]. We used both of these genes to test the possibility of designing a Mammalia Chip. We also used these sequences and various size fragments of them to test the feasibility of DNA barcoding analysis for mammalian species. We targeted three datasets of mammalian species for these analyses: 121 species across the taxonomy of mammals (mammalian dataset), a dense sampling of 87 species of neotropical bats (bat dataset), and a wide geographical sampling of a single genetically diverse bat species (Sturnira lilium dataset).

This study reveals that both arrays and barcodes are useful tools for the species-level identification of mammals. The main limitation of the array-based approach is that it requires advance knowledge of sequences in target species. Because of a lack of exact matches, undiscovered haplotypes or geographic variants could fail to anneal properly to the probes on the array. While we tried to avoid this problem by providing a set of three different probes per species, this factor can substantially limit the use of microarrays for large-scale biodiversity monitoring. Additionally, to explore unknown species in a given taxonomic group, it might be possible to design probe sets that specifically bind to members of a higher taxonomic level such as genus or family. However, in a situation such as S. lilium, with different geographical variants of up to 8% sequence variation in their COI/cytb genes, a probe set that is designed for the species in one locality might not bind to members of the species in other localities (results not shown). This hit or miss situation could make array technology less desirable in biodiversity monitoring across a wide geographic region. In fact, the current applications of microarrays are usually focused on a limited number of taxa [4]. Because of this, assembling a 'Mammalia Chip' might not be a feasible approach for biodiversity monitoring of all mammalian species.

Because barcoding is a sequencing-based technology, it avoids the problem of unknown haplotypes. New haplotypes can be compared to existing databases of barcodes, and they can be assigned to a particular species using probabilistic algorithms [15, 16]. The final assignment of a new haplotype to a described species or its assignment to a new species will be achieved through comprehensive taxonomic analysis, which requires different types of data [17]. Our analysis supports this argument in all three datasets. While smaller fragments were less powerful in resolving some closely-related species, obtaining more sequence information in these cases (i.e. full-length barcode versus mini-barcode or the whole gene versus the barcode-size fragment) can increase the resolution [10]. However, while standard barcode-size fragments (650 bp) can be readily obtained in a single PCR amplification/sequencing from freshly collected or frozen tissue specimens, it is difficult to obtain 650-bp barcodes from specimens whose DNA is degraded (i.e. dried museum samples) [10]. The high effectiveness of mini-barcodes means that biomonitoring through barcodes can target different types of specimens, including museum samples or traces of tissues with degraded DNA [10]. The mini-barcode strategy also enables exploration of the use of massively parallel sequencing platforms, such as pyrosequencing-based [18] 454 Life Sciences sequencers, for barcoding applications. Interestingly, this technology uses an emulsion PCR approach for simultaneous amplification of several thousand 100–200 base DNA molecules in one reaction. This approach will therefore allow the use of mini-barcodes on environmental samples, which have traditionally been targets for array-based technology.

This study also provides evidence that both COI and cytb are useful species-level molecular markers for mammalian species. This finding is in agreement with earlier work [4]. However, when it comes to selecting a molecular marker, it is also important to consider operational issues such as the availability of robust PCR primers, standardization across a wide range of taxa, the robustness of amplifying shorter fragments in PCR reactions of degraded DNA, and the prevalence of mitochondrial nuclear pseudogenes. Our study further confirms that the standard COI barcode can be applied to mammalian species with a similar high species-level resolution as has been observed in other animal taxa tested.

DNA-based methods such as DNA arrays and DNA barcodes provide substantial potential for biodiversity monitoring. However, as the scale of analysis increases, for example in large biodiversity surveys or analysis across wide taxonomic assemblages or different types of specimens, the scalability and sensitivity of these approaches become critical issues in their applicability. Our analyses using three different datasets of mammalian species spanning a wide range of taxa, suggest both DNA arrays and DNA barcodes provide high resolution (i.e. ~95%) across mammalian species. Because DNA arrays might fail to anneal to undiscovered haplotypes of a given species, their use is limited to taxa with known sequences. DNA barcoding, however, provides a higher flexibility for the identification of species in large taxonomic assemblages because it is based on obtaining sequence information that can be used for linking unknown haplotypes to known species.