DNA barcoding promises to revolutionize the way taxonomists work, facilitating species identification by using small, standardized portions of the genome as substitutes for morphology. The concept has gained considerable momentum in many animal groups, but the higher plant world has been largely recalcitrant to the effort. In plants, efforts are concentrated on various regions of the plastid genome, but no agreement exists as to what kinds of regions are ideal, though most researchers agree that more than one region is necessary. One reason for this discrepancy is differences in the tests that are used to evaluate the performance of the proposed regions. Most tests have been made in a floristic setting, where the genetic distance and therefore the level of variation of the regions between taxa is large, or in a limited set of congeneric species.
Methodology and Principal Findings
Here we present the first in-depth coverage of a large taxonomic group, all 86 known species (except two doubtful ones) of crocus. Even six average-sized barcode regions do not identify all crocus species. This is currently an unrealistic burden in a barcode context. Whereas most proposed regions work well in a floristic context, the majority will – as is the case in crocus – undoubtedly be less efficient in a taxonomic setting. However, a reasonable but less than perfect level of identification may be reached – even in a taxonomic context.
The time is ripe for selecting barcode regions in plants, and for prudent examination of their utility. Thus, there is no reason for the plant community to hold back the barcoding effort by continued search for the Holy Grail. We must acknowledge that an emerging system will be far from perfect, fraught with problems and work best in a floristic setting.
Citation: Seberg O, Petersen G (2009) How Many Loci Does it Take to DNA Barcode a Crocus? PLoS ONE 4(2): e4598. https://doi.org/10.1371/journal.pone.0004598
Editor: Jane Catherine Stout, Trinity College Dublin, Ireland
Received: November 12, 2008; Accepted: January 12, 2009; Published: February 25, 2009
Copyright: © 2009 Seberg et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: This research was supported by the Danish Natural Sciences Research Council (272-06-0436), the Sloan Foundation, and Gordon and Betty Moore Foundation. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
The primary aims of DNA barcoding are to identify known specimens and to help flag possible new species, thereby making taxonomy more effective for science and society (www.barcoding.si.org).
The majority of DNA barcoding studies in animals have used a small region of the mitochondrial gene, cytochrome c oxidase subunit I (COI), as barcode, and COI is promising, with a few exceptions , to be the universal barcode for animals . The search for a universal barcode in plants has been far more tortuous, but general agreement is emerging that more than one region is needed – (for a dissenting view; [see 8]), and that these regions will need to come from the plastid genome, where several different regions have been suggested. The plant mitochondrial genome is usually far too conservative to be of use –, , whereas the plastid genome shares several of the desirable properties found in the animal mitochondrion. The nuclear multi-copy internal transcribed spacers array (ITS) has also been suggested as a barcode , , but has been discarded due to its peculiar pattern of evolution  and currently the nuclear genome is largely inaccessible for barcoding purposes. However, current advances in sequencing technology and the diminishing expenses promise radically to change the way we do barcoding; [see e.g. 11]. The majority of plastid regions have been proposed on the basis of comparisons of levels of variation in whole genomes of closely related taxa, e.g. congeneric species, and subsequently tested in taxonomically widely dispersed, randomly chosen species pairs , ,  and/or in a purely floristic setting , . As a consequence, emphasis is placed on primer universality at the expense of species recognition; see however Fazekas et al. .
Chase et al. , using data predominantly drawn from GenBank, and acknowledging the imperfect nature of these data, suggested the use of one or two, unspecified, plastid regions plus ITS as potentially useful plant barcodes. Almost simultaneously Kress et al.  pointed to the potential value of a combination of ITS and the trnH-psbA spacer. The choice of the latter region was based on a combination of comparisons of the complete plastid genomes of deadly nightshade (Atropa belladonna L.) and tobacco (Nicotiana tabacum L.), and tests on a small, taxonomically defined set of species and a larger one defined geographically. Using a similar approach, but restricting the search to coding plastid regions only, Chase et al.  selected a number of potentially useful regions (see www.kew.org/barcoding) that subsequently were tested by a number of research groups worldwide on a limited taxon set. Based on an evaluation of their overall performance in these tests, a smaller set was chosen subsequently, and these favoured regions tested more widely, and two sets were proposed eventually as universal barcodes, rpoC1 and matK plus either rpoB or trnH-psbA. However, these two triplets were challenged by Kim et al., who preferred a combination of matK, atpF-H, and psbI-psbK .
On the basis of its ability to distinguish between congeneric species-pairs, Newmaster et al.  proposed rbcL as the first tier in their proposal of a two-tiered approach to barcoding; allowing the next region to be an optional choice. Kress and Erickson  refined this proposal by suggesting only the 3′ end of rbcL as their first choice and adding trnH-psbA as their preferred second choice, again basing their conclusion on tests involving congeneric species-pairs. However, using a relatively dense taxonomic coverage (35% (mobot.mobot.org/W3T/Search/vast.html) of known species in a single genus), Newmaster et al. , while maintaining their choice of trnH-psbA, replaced rbcL with matK. This conclusion was supported by Lahaye et al.  on the basis of geographically defined, purely floristic and similarly designed but broader taxonomic studies in a test of all regions suggested by Chase et al. .
Acknowledging the low resolution of the trnL intron, even in a floristic context, Taberlet et al.  none the less suggested its use as a plant barcode, primarily due to the capability of its P6 loop to distinguish species in highly degraded or processed material. Based on a taxonomic study with the hitherto densest taxon sampling, 48% of all 278 known species of a single genus, Edwards et al.  suggested that at least three regions were necessary to discriminate most of the studied species, including a new region trnT-trnL in combination with ITS and “at least one more region with a greater level of variation than psbA-trnH” .
Results and Discussion
Here we present the first analysis of the performance of one of the recently proposed barcode sets by Chase et al. , rpoC1, matK, and trnH-psbA, in a large, taxonomically defined, monophyletic group; the genus Crocus L. (Iridaceae). We tested these regions on 86 (98%) species of the genus Crocus, excluding only two species of doubtful taxonomic status. To this set of regions we have added three other regions, two of which have been considered of potential value as barcodes (rps8-rpl36  and accD ), plus ndhF, which we have used for phylogenetic purposes . The taxon sampling was not designed to capture intraspecific variation, but 17 species include more than one accession each (from 2 to 15). Though few studies of plants take intraspecific variation into account, this may be a bigger problem than generally believed, by reducing the barcode gap and consequently the success rate of identification [see 7].
The proposed barcode set  is diagnostic for 63 (73%) of the included Crocus-species, which is only marginally better than the combination matK and psbA-trnH alone, which identifies 62 (72%) species (Table 1). Substituting rpoC1 with any of the two other considered regions, accD or rps8-rpl36, improves species recognition to 67 (78%) and 65 (76%), respectively. Using the four most variable of these above-mentioned regions (ndhF, matK, trnH-psbA, and rps8-rpl36) makes it is possible to identify 79 species (92%), a figure that is not changed by adding the two least variable regions (accD and rpoC1). In all instances the species level resolution is higher than the ones obtained by Fazekas el al.  using a similar number of loci. This is undoubtedly due to our non-tree based approach, which does not require gene-tree monophyly . Interestingly, ndhF, which has not been suggested as a barcode due to lack of primer universality, has a higher resolving power than matK, which is one of the top candidates as a universal barcode. However, in most instances some regions will perform better than others no matter which are chosen as barcodes . In general, identification success was not influenced by the inclusion of length variation. It is worth mentioning that the two alternative regions proposed by Kim et al. at the Consortium for the Barcode of Life meeting in Taipei, in addition to matK, do not behave, in a smaller subset of Crocus species (see Tables S1 and S2), better than any of those suggested by Chase et al. .
With the notable exception of rbcL, most of the plastid regions that have been suggested as official barcodes, and hence potentially being the most variable regions, stems from the same limited region, covering approximately ∼15% of the large single-copy region of the plastid, spanning from rpoB to trnH-psbA. Hence, it is not surprising that “there are multiple multilocus plant DNA barcoding combinations that perform about equally well in resolving species” .
In figure 1 the relationship between the number of Crocus species identified is plotted as a function of the number of basepairs sequenced and the best fitting curve is added. Solving the equation for y = 86 (the maximum number of species) gives x = 5859, and it appears reasonable, at least in theory to postulate that it would require approximately 5800 bp from the plastid genome to identify all known species of Crocus. This corresponds to 9–10 average-sized (∼600 bp) barcode genes/regions and is presently not a workable option. Using a differently defined taxon sampling and a different criterion for correct identification a similar relationship between number of barcode regions and discrimination ability was found by Fazekas et al. .
The genus includes 86 known species, and the six regions (five proposed barcode regions plus ndhF) used here identify 79 species. The regions are added according to their ability to identify species. The performance of individual genes is shown in Table 2. The logarithmic trend line (y = 12.3 ln(x)−21.0) and the R2 ( = 0.97) were calculated in Excel® and checked in JMP, Version 7. SAS Institute Inc., Cary, NC, 1989–2007.
Also, it would be problematic to flag potentially new species using these data because several species, most notably C. biflorus Mill. and C. reticulates Steven ex Adams, have non-monophyletic sets of plastid genomes (see figure S1). Disagreement between the evolutionary histories of organellar and nuclear genomes are not uncommon in plants ([see e.g. 19]) and animals , and it is a controversial point whether reciprocal monophyly  is a necessary requirement in barcoding. However, if reciprocal monophyletic species are mandatory this is bound to decrease identification success; see e.g. Fazekas et al. . Using the same approach as Fazekas et al.  and a bootstrap cut-off value of 70% only two of the 17 species represented by more than one accession in the present analysis are monophyletic and two paraphyletic.
In a barcoding context (e.g. Barcode of Life Data System (BOLD); www.boldsystems.org ), identification is often conducted by reference to clusters of taxa in a neighbor-joining (NJ) tree using an average Kimura-2-parameter model. If this approach is used on the Crocus data set, many of the branch lengths are extremely short, making species assignment of new accessions spurious at best (figure 2). To aggravate this, a whole suite of NJ trees may be produced from the same dataset, due to the known input-order sensitivity of NJ. Resampling techniques are occasionally used to justify species monophyly [see e.g. 7], ,  or to justify species identification; see e.g. Fazekas et al. . Due to their innate properties } and their very different implementation in different computer programmes  the use of resampling in this context is ill-advised, and would have made little sense here.
The extremely short branches make identification by cluster memberships difficult as does the “non-monophyly” of several species. Species that cannot be identified by any of the sequences used are marked in blue. The NJ tree is rooted with the two Crocus species that are sister group to the remaining Crocus species in the parsimony based phylogeny (see figure S1).
As in any other group of species, species circumscription in Crocus may be problematic, and this may be one factor underlying the observed lack of resolution. However, there is no reason to believe that taxonomic problems are more severe in Crocus than in many other taxa. Even in groups with an almost universally agreed upon species-level taxonomy, e.g., barley (Hordeum L.) with 32 species, it is impossible to recognise more than approximately 50% of the species using matK and rpoC1 (see Table S3). Perhaps the worst case scenario is found among the morphologically distinct species of the Galápagos sunflower tree, Scalesia Arn.(Asteraceae), where no variation has been found (see Table S4) in the plastid markers, and almost none in nuclear markers.
Barcode efforts in plants are severely hampered by an obvious lack of agreement about the choice of regions. Opinions are divided on whether two or three regions suffice, whether regions should be coding or non-coding, whether one should or should not use a tiered approach, and ultimately on which regions are to be preferred and how the data are analysed, and the baseline for comparison has been ill-defined , –. In the most thorough comparison of all suggested regions to date, both with respect to technical and variability issues, it is recommended that three (or even four) regions, one coding (rbcL, rpoB, matK) and two non-coding (trnH-psbA, atpF-atpH), are selected as official barcodes  as this represents a reasonable compromise between effort and species resolution as it is currently implemented in the Barcode of Life Data System.
Both the present study and the study by Fazekas et al.  show that there is a limit to resolution no matter which region or regions are chosen. The present study shows that in a taxonomic setting and with a reasonable effort it is unlikely that barcoding will enable us to identify more than around 70–75% of the known species – in some instances less, in others more.
However, the time is ripe for selecting barcode regions in plants, and for prudent examination of their utility. Based on the level of sequence variation alone, an optimal set of regions is not yet known, but matK and trnH –psbA are strong candidates, though other conditions much notably primer universality and sequence quality have to be taken into account. However, we must acknowledge that the emerging system will be far from perfect ([see e.g. 29]), and that it will work best in a floristic setting.
Thus, there is no reason for the plant community to hold back the barcoding effort by continued search for the Holy Grail .
Materials and Methods
Taxon sampling was as extensive as possible. Of the 81 species recognized by Mathew  all but one, Crocus boissieri Maw, known only from a herbarium specimen, are included, as are six of the seven species described since then. Of the 50 recognized subspecies  48 are included, but only two of the 10 later described ones. A total of 17 species were represented by more than one accession, and the total number of included accessions of Crocus is 131. Two species of Romulea Maratti, and one species each of Syringodea D. Don, Babiana Sims, and Tigridia Juss. were included as outgroups. Voucher information and GenBank accession numbers may be found in Petersen et al. .
DNA extractions were performed using the DNeasy Plant Mini Kit (QIAGEN Ltd., Crawley, West Sussex) after tissue disruption in a FastPrep FP-120 bead mill (Qbiogene, Carlsbad, CA). PCR amplifications followed standard procedures except for the addition of 0.1 mg/ml BSA to most reactions. For PCR amplification and sequencing of the five plastid regions the following primers were used: ndhF1318F and ndhF2110R , accD1F and accD3R (http://www.kew.org/barcoding/protocols.html), rpoC1F and rpoC4R (http://www.kew.org/barcoding/protocols.html), rpl36F and rps8R , and psbAF and trnH2 , . Direct sequencing of purified PCR products was performed using BIGDYE 1.1 (Applied Biosystems, Wellesley, Massachusetts, USA) and purified sequencing products were run on an AB3130xl automated sequencer (Applied Biosystems). Sequence editing was done using Sequencher versions 4.5 to 4.7 (Gene Codes Corporation, Ann Arbor, Michigan, USA).
In order to assess the potential value of the six sequence regions as barcodes, the outgroup taxa and the hybrid taxon C.×jessopae Bowles were excluded. All variable sites were included, and so was an ambiguously aligned region of trnH-psbA previously excluded from phylogenetic analysis. PAUP*, version 4.0b8  provided numbers of variable sites. The number of uniquely identifiable species was checked using MacClade version 4.08 . A species is considered uniquely identifiable if all the included specimens/subspecific taxa can be identified. Thus, species monophyly is not a requirement. Alignments were done manually and the matrix is available at TreeBase (acc no. M3519, S1912).
All phylogenetic analyses were performed using PAUP*, version 4.0b8 . Uninformative characters were excluded from the phylogenetic analyses, and informative characters were equally weighted and treated as unordered. Gaps were treated as ambiguous data (?). Analyses were performed using both the default branch collapsing rule (collapse if maximum is zero) and amb- (collapse if minimum length is zero). The latter option was used for facilitating comparison of results from phylogenetic analyses using PAUP* with result from analyses using WinClada . Under the default branch collapsing rule and simple sequence addition the number of equally parsimonious trees was very high (hitting the limit of 637.000 defined by memory allocation) and analyses without an upper limit for the number of saved trees could not be run to completion. Thus, we also used a two step approach first running 1.000 random addition sequences saving no more than 25 trees per replicate. The trees saved in this analysis were used as starting trees for a new analysis with a maximum number of trees saved set to 100.000. Phylogenetic analyses performed using WinClada, version 1.00.08 , spawning the matrix to NONA version 2.0  were executed using heuristic search options hold10000, mult*100, max*, hold/10, and the default branch collapsing rule, amb-.
Neighbor-joining was also done in PAUP 4.0b8, using the default settings and Kimura-2-parameter distance option.
Strict consensus tree of Crocus and five outgroup taxa.
(6.33 MB TIF)
Sequence variation and species identification ability of eight plastid regions in Crocus series Crocus. Crocus series Crocus is monophyletic (see figure S1) and includes nine species (C. sativus L., C. cartwrightianus Herb., C. hadriaticus Herb., C. thomasii Ten., C. oreocreticus B. L. Burtt, C. asumaniae B. Mathew & T. Baytop, C. mathewii Kernd. & Pasche, C. pallasii Goldb., C. moabiticus Bornm. & Dism. ex Bornm). Three species (C. sativus, C. cartwrightianus, C. hadriaticus) cannot be identified be any sequence. The length of the region atpF-H is 570–572 bp (573 bp in alignment). atpF-H GenBank acc. nos. EU523361-EU523373. The region psbI-K is very short (ca. 173–179 bp), but difficult to sequence due to several longer runs of T's (at least 3 runs of 9–10 or more T's).
(0.03 MB DOC)
Sequence variation and species identification ability of six plastid regions in various combinations in Crocus serie Crocus. See Table S1
(0.03 MB DOC)
Sequence variation and species identification ability of two plastid regions in Hordeum (all 32 species). No length variation is observed among the sequences. GenBank acc. nos. EU118371-EU118422, EU118427-EU118478.
(0.04 MB DOC)
Sequence variation and species identification ability of six plastid regions in Scalesia (5 of 15 species). No length variation is observed among the sequences. GenBank acc. nos. EU118423-EU118426, EU118479-EU118483, EU118494-EU118498, EU118509-EU118513, EU118524-EU118527, EU118536-EU118539.
(0.04 MB DOC)
Jerry I. Davis, Mark Stoeckle, Dennis Wm. Stevenson, Damon Little, and Finn Borchsenius are thanked for comments that helped improve the manuscript.
Conceived and designed the experiments: OS GP. Performed the experiments: OS GP. Analyzed the data: OS GP. Contributed reagents/materials/analysis tools: OS GP. Wrote the paper: OS GP.
- 1. Blaxter ML (2004) The promise of a DNA taxonomy. Philo Trans R Soc Lond B Biol Sci 359: 669–679.ML Blaxter2004The promise of a DNA taxonomy.Philo Trans R Soc Lond B Biol Sci359669679
- 2. Waugh J (2007) DNA barcoding in animal species: progress, potential and pitfalls. Bioessays 29: 188–197.J. Waugh2007DNA barcoding in animal species: progress, potential and pitfalls.Bioessays29188197
- 3. Chase MW, Salamin N, Wilkinson M, Dunwell JM, Kesanakurthi RP, et al. (2005) Land plants and DNA barcodes: short-term and long-term goals. Philos Trans R Soc Lond B Biol Sci 360: 1889–1895.MW ChaseN. SalaminM. WilkinsonJM DunwellRP Kesanakurthi2005Land plants and DNA barcodes: short-term and long-term goals.Philos Trans R Soc Lond B Biol Sci36018891895
- 4. Kress WJ, Wurdack KJ, Zimmer EA, Weigt LA, Janzen DH (2005) Use of DNA barcodes to identify flowering plants. Proc Natl Acad Sci USA 102: 8369–8374.WJ KressKJ WurdackEA ZimmerLA WeigtDH Janzen2005Use of DNA barcodes to identify flowering plants.Proc Natl Acad Sci USA10283698374
- 5. Newmaster SG, Fazekas AJ, Ragupathy S (2006) DNA barcoding in land plants: evaluation of rbcL in a multigene tiered approach. Can J Bot 84: 335–341.SG NewmasterAJ FazekasS. Ragupathy2006DNA barcoding in land plants: evaluation of rbcL in a multigene tiered approach.Can J Bot84335341
- 6. Cowan RS, Chase MW, Kress WJ, Savolainen V (2006) 300,000 species to identify: problems, progress, and prospects in DNA barcoding of land plants. Taxon 55: 611–616.RS CowanMW ChaseWJ KressV. Savolainen2006300,000 species to identify: problems, progress, and prospects in DNA barcoding of land plants.Taxon55611616
- 7. Fazekas AJ, Burgess KS, Kesanakurti PR, Graham SW, Newmaster SG, et al. (2008) Multiple Multilocus DNA Barcodes from the Plastid Genome Discriminate Plant Species Equally Well. PLoS ONE 3: e2802.AJ FazekasKS BurgessPR KesanakurtiSW GrahamSG Newmaster2008Multiple Multilocus DNA Barcodes from the Plastid Genome Discriminate Plant Species Equally Well.PLoS ONE3e2802
- 8. Lahaye R, van der Bank M, Bogarin D, Warner J, Pupulin F, et al. (2008) DNA barcoding the floras of biodiversity hotspots. Proc Natl Acad Sci U S A 26: 2923–2928.R. LahayeM. van der BankD. BogarinJ. WarnerF. Pupulin2008DNA barcoding the floras of biodiversity hotspots.Proc Natl Acad Sci U S A2629232928
- 9. Muse SV (2000) Examining rates and patterns of nucleotide substitution in plants. Plant Mol Biol 42: 25–43.SV Muse2000Examining rates and patterns of nucleotide substitution in plants.Plant Mol Biol422543
- 10. Wendel JF, Schnabel A, Seelanan T (1995) Bidirctional interlocus concereted eolution following allopolyploid speciciation in cotton (Gossypium). Proc Natl Acad Sci U S A 92: 280–284.JF WendelA. SchnabelT. Seelanan1995Bidirctional interlocus concereted eolution following allopolyploid speciciation in cotton (Gossypium).Proc Natl Acad Sci U S A92280284
- 11. Kane NC, Cronk QCB (2008) Meeting Review: Botany without borders: barcoding in focus. Mol Ecol 17: 5175–5176.NC KaneQCB Cronk2008Meeting Review: Botany without borders: barcoding in focus.Mol Ecol1751755176
- 12. Kress WJ, Erickson DL (2007) A Two-Locus Global DNA Barcode for Land Plants: The Coding rbcL Gene Complements the Non-Coding trnH-psbA Spacer Region. PLoS ONE 2: e508.WJ KressDL Erickson2007A Two-Locus Global DNA Barcode for Land Plants: The Coding rbcL Gene Complements the Non-Coding trnH-psbA Spacer Region.PLoS ONE2e508
- 13. Taberlet P, Coissac E, Pompanon F, Gielly L, Miquel C, et al. (2007) Power and limitations of the chloroplast trnL (UAA) intron for plant DNA barcoding. Nucleic Acids Res 35: e14.P. TaberletE. CoissacF. PompanonL. GiellyC. Miquel2007Power and limitations of the chloroplast trnL (UAA) intron for plant DNA barcoding.Nucleic Acids Res35e14
- 14. Chase MW, Cowan RS, Hollingsworth PM, van den Berg C, Madriñán S, et al. (2007) A proposal for a standarised protocol to barcode all land plants. Taxon 56: 295–299.MW ChaseRS CowanPM HollingsworthC. van den BergS. Madriñán2007A proposal for a standarised protocol to barcode all land plants.Taxon56295299
- 15. Pennisi E (2007) Wanted: A barcode for plants. Science 318: 190–191.E. Pennisi2007Wanted: A barcode for plants.Science318190191
- 16. Edwards D, Horn A, Taylor D, Savolainen V, Hawkins JA (2008) DNA barcoding of a large genus, Aspalathus L. (Fabaceae). Taxon 57: 1317–1327.D. EdwardsA. HornD. TaylorV. SavolainenJA Hawkins2008DNA barcoding of a large genus, Aspalathus L. (Fabaceae).Taxon5713171327
- 17. Presting GG (2006) Identification of conserved regions in the plastid genome: implications for DNA barcoding and biological function. Can J Bot 84: 1434–1443.GG Presting2006Identification of conserved regions in the plastid genome: implications for DNA barcoding and biological function.Can J Bot8414341443
- 18. Petersen G, Seberg O, Thorsø S, Jørgensen T, Mathew B (2008) A phylogeny of the genus Crocus (Iridaceae) based on sequence data from five plastid regions. Taxon 57: 487–499.G. PetersenO. SebergS. ThorsøT. JørgensenB. Mathew2008A phylogeny of the genus Crocus (Iridaceae) based on sequence data from five plastid regions.Taxon57487499
- 19. Syring J, Farrell K, Businsky R, Cronn R, Liston A (2007) Widespread genealogical nonmonophyly in species of Pinus subgenus Strobus. Syst Biol 56: 163–181.J. SyringK. FarrellR. BusinskyR. CronnA. Liston2007Widespread genealogical nonmonophyly in species of Pinus subgenus Strobus.Syst Biol56163181
- 20. Funk DJ, Omland KE (2003) Species-Level Paraphyly and Polyphyly: Frequency, Causes, and Consequences, with Insights from Animal Mitochondrial DNA. Annu Rev Ecol Syst 34: 397–423.DJ FunkKE Omland2003Species-Level Paraphyly and Polyphyly: Frequency, Causes, and Consequences, with Insights from Animal Mitochondrial DNA.Annu Rev Ecol Syst34397423
- 21. Meyer CP, Paulay G (2005) DNA Barcoding: Error Rates Based on Comprehensive Sampling. PLoS Biology 3: 2229–2238.CP MeyerG. Paulay2005DNA Barcoding: Error Rates Based on Comprehensive Sampling.PLoS Biology322292238
- 22. Costa FO, DeWaard JR, Boutillier J, Ratnasingham S, Dooh RT, et al. (2007) Biological identifications through DNA barcodes: the case of the Crustacea. Can J Fish Aquat Sci 64: 272–295.FO CostaJR DeWaardJ. BoutillierS. RatnasinghamRT Dooh2007Biological identifications through DNA barcodes: the case of the Crustacea.Can J Fish Aquat Sci64272295
- 23. Hajibabaei M, Janzen DH, Burns JM, Hallwachs W, Hebert PDN (2006) DNA barcodes distinguish species of tropical Lepidoptera. Proc Natl Acad Sci USA 103: 968–971.M. HajibabaeiDH JanzenJM BurnsW. HallwachsPDN Hebert2006DNA barcodes distinguish species of tropical Lepidoptera.Proc Natl Acad Sci USA103968971
- 24. Hillis DM, Bull JJ (1993) An Empirical Test of Bootstrapping as a Method for Assessing Confidence in Phylogenetic Analysis. Syst Biol 42: 182–192.DM HillisJJ Bull1993An Empirical Test of Bootstrapping as a Method for Assessing Confidence in Phylogenetic Analysis.Syst Biol42182192
- 25. Davies TJ, Barraclough TG, Chase MW, Soltis PS, Soltis DE, et al. (2004) Darwin's abominable mystery: Insights from a supertree of the angiosperms. Proc Natl Acad Sci USA 101: 1904–1909.TJ DaviesTG BarracloughMW ChasePS SoltisDE Soltis2004Darwin's abominable mystery: Insights from a supertree of the angiosperms.Proc Natl Acad Sci USA10119041909
- 26. Helfenbein KG, DeSalle R (2005) Falsifications and corroborations: Karl Popper's influence on systematics. Mol Phylogenet Evol 35: 271–280.KG HelfenbeinR. DeSalle2005Falsifications and corroborations: Karl Popper's influence on systematics.Mol Phylogenet Evol35271280
- 27. Little DP, Stevenson DW (2007) A comparison of algorithms for the identification of specimens using DNA barcodes: examples from gymnosperms. Cladistics 23: 1–21.DP LittleDW Stevenson2007A comparison of algorithms for the identification of specimens using DNA barcodes: examples from gymnosperms.Cladistics23121
- 28. Kress WJ, Erickson DL (2008) DNA barcodes: Genes, genomics, and bioinformatics. Proc Natl Acad Sci USA 105: 2761–2762.WJ KressDL Erickson2008DNA barcodes: Genes, genomics, and bioinformatics.Proc Natl Acad Sci USA10527612762
- 29. Sass C, Little DP, Stevenson DW, Specht CD (2007) DNA barcoding in the cycadales: testing the potential of proposed barcoding markers for species identification of cycads. PLoS ONE 2: e1154.C. SassDP LittleDW StevensonCD Specht2007DNA barcoding in the cycadales: testing the potential of proposed barcoding markers for species identification of cycads.PLoS ONE2e1154
- 30. Rubinoff D, Cameron S, Will K (2006) Are plant DNA barcodes a search for the Holy Grail? Trends Ecol Evol 21: 1–2.D. RubinoffS. CameronK. Will2006Are plant DNA barcodes a search for the Holy Grail?Trends Ecol Evol2112
- 31. Mathew B (1982) The Crocus. Portland, OR: Timber Press. B. Mathew1982The CrocusPortland, ORTimber Press127
- 32. Olmstead RG, Sweere JA (1994) Combining data in phylogenetic systematics: an empirical approach using three molecular data sets in the Solanaceae. Syst Biol 43: 467–481.RG OlmsteadJA Sweere1994Combining data in phylogenetic systematics: an empirical approach using three molecular data sets in the Solanaceae.Syst Biol43467481
- 33. Kress WJ, Liu AZ, Newman M, Li QJ (2005) The molecular phylogeny of Alpinia (Zingiberaceae): A complex and polyphyletic genus of gingers. Amer J Bot 92: 167–178.WJ KressAZ LiuM. NewmanQJ Li2005The molecular phylogeny of Alpinia (Zingiberaceae): A complex and polyphyletic genus of gingers.Amer J Bot92167178
- 34. Sang T, Donoghue MJ, Zhang D (1997) Evolution of alcohol dehydrogenase genes in peonies (Paeonia); phylogenetic relationships of putative nonhybrid species. Mol Biol Evol 14: 994–1007.T. SangMJ DonoghueD. Zhang1997Evolution of alcohol dehydrogenase genes in peonies (Paeonia); phylogenetic relationships of putative nonhybrid species.Mol Biol Evol149941007
- 35. Tate JA, Simpson BB (2003) Paraphyly of Tarasa (Malvaceae) and diverse origins of the polyploidy species. Syst Bot 28: 723–737.JA TateBB Simpson2003Paraphyly of Tarasa (Malvaceae) and diverse origins of the polyploidy species.Syst Bot28723737
- 36. Swofford DL (2001) PAUP*. Phylogenetic Analysis Using Parsimony (* and Other Methods), version 4.0b8 [computer program]. Sunderland, Mass.: Sinauer Associates. DL Swofford2001PAUP*. Phylogenetic Analysis Using Parsimony (* and Other Methods), version 4.0b8 [computer program]Sunderland, Mass.Sinauer Associates
- 37. Maddison WP, Maddison DR (2000) MacClade, version 4.08 [computer program]. Sunderland, Massachusetts: Sinauer Associates. WP MaddisonDR Maddison2000MacClade, version 4.08 [computer program]Sunderland, MassachusettsSinauer Associates
- 38. Nixon KC (2002) WinClada, version 1.00.08 [computer program]. Ithaca, NY.: The Author. KC Nixon2002WinClada, version 1.00.08 [computer program]Ithaca, NY.The Author
- 39. Goloboff PA (1999) NONA (NO NAME), version 2 [computer program]. Tucumán, Argentina: The Author. PA Goloboff1999NONA (NO NAME), version 2 [computer program]Tucumán, ArgentinaThe Author