There are many reasons why morphological taxonomy can't be replaced by DNA barcoding. Here are a few that I can think of off the top of my head.

(1) Introgression, where whole genes can pass unchanged between species, will lead to misleading identifications.

(2) Nuclear mitochondrial pseudogenes (NUMTs) result in imperfect identifications.

(3) You need a full referenced database in the first place to do DNA barcoding, which requires traditional taxonomy.

(4) A single gene (i.e. COI) may not be representative of the whole genome, and might be unchanged in recently diverged species. While the potential incongruency of COI (particularly with regard to genetic saturation) with other loci and the species phylogeny and the idea that “one gene fits all” might lead to unrepresentative biodiversity patterns

(5) The practicality of amplifying and sequencing universal stretches of DNA is not universal.

(6) The identification of species, via DNA barcoding, relies on the idea that their sequences will cluster so that intraspecific divergences will always be less than interspecific divergences (the barcode gap), and that this gap is consistent and diagnosable across all animal species (this is not always true).

(7) Incomplete lineage sorting leads to barcode gap overlap, which lumps overwise distinct morphological species. Diagnosable morphological differences that develop due to divergent selection are likely to manifest much sooner than a neutral gene, therefore species can be detected much sooner than variation at neutral markers.

(8) Even if a barcode gap exists, it is unlikely to be consistent among lineages. Rates of evolution differ among lineages as well as genes.

I feel that DNA barcoding is an amazing tool for the rapid identification of species, but its use in the absence of traditional taxonomy will be unreliable. DNA barcoding and traditional morphological taxonomy are two tools that taxonomists should use in an integrative, iterative way. Abandoning one for the other would be a step backwards.

To add a short answer: DNA barcoding is not intended to be used to answer questions in taxonomy or systematics, you cannot obtained sound species trees with a single, maternally inherited marker. Biut it providesa simple and fast "quick sorting system". Highly similar or identical barcodes make it at least very likely that samples belong to the same taxon (often: "species"). Without the need to have any mophological identification skills this is a great way for sorting samples, doing ecological assessments or monitoring. That's it i.m.o.

DNA is not available for many species, especially those known only as fossils or extinct or rare species known only through museum specimens preserved by methods incompatible with DNA analysis.

I can't bank on the sentence that "DNA barcoding is an amazing tool for the rapid identification of species", I would rather define it as a tool to discriminate among taxonomic units. I can add that a) it does not help when dealing with description of new species to science and b) databases include lots of misidentifications of voucher specimens from where sequences have been sampled and labeled. 

Unionid freshwater bivalves, Mytilidae and some Veneridae exhibit Double Uniparental Inheritance, that is they have both male and female mitochondrial DNA lineages. if male and female specimens are being compared there will be some amazing problems in analyses. Likewise some of the posted sequences are not labeled as to the sex of the animal sequenced. Using bar codes can be a useful tool but will not replace systematics

To simplify Cuong's (and other) excellent response, there are basically two reasons why DNA barcoding cannot replace traditional taxonomy: 1) we are far from having a complete database with all animal COI sequences (and far from knowing all animals out there); and 2) even if we had that, the barcoding gene cannot be used reliably to identify many species. You should think of DNA sequencing as an additional tool in taxonomy, not the only one.

Indeed you can reduce the excellent answers above further.  You need to have the taxonomy in the first place before you can tie a barcode marker to a taxon.  In an ideal world the barcode marker would be taken from the type, as the type ties the genetic or biological material to a name.  However, currently our technological ability to do this for old material is limited. So in practical terms you need a good voucher so if there is any doubt or questions arise (as we see in many published datasets) you can go back to the voucher and confirm the identification.

As an aside while this conversation is about animal groups and COI which has proven to be an excellent barcode region for animals in the plant workld we suffer from the lack of a universal barcode that can be applied across the kingdom.

In addition to the other excellent remarks I would like to add that DNA barcoding has been based on the species as they were defined (in many cases anyway) by traditional taxonomy. In itself that excludes replacing it already. Then there is also the underlying species concept. If we were to use DNA barcoding then we would start to use a concept of genetic distances based on one mtDNA fragment. It has been shown often enough that this simply does not hold.

There is one main reason why not

What we actually want to find out in taxonomy, is a phylogenetic signal, which allows us to delimit the units which have independent evolutionary histories (due to speciation events).

We use the observed variation in different character systems as proxies for this task.

The best practice is to use as many aspects of biodiversity available, to avoid misleading effects present in any single character system used as a proxy.

So an integrative approach is preferable, regardless on which system we assess to be most reliable. We do not delimit species solely on the basis of characters of the genitalia in all organisms. The same is true for DNA-Barcodes.

There might be different views of how reliable DNA Barcoding is in a given group of organisms, but adding as much information as possible to drive a conclusion (be it taxonomic or from any different field of biology) is the best way to ensure high plausibility and intersubjective repeatabilty of any conclusion.

Thank You All

To complete the very good analysis of Cuong

If your species are younger than 3 million years.....barcoding (at least on Cox1) is not working...you will not have any resolution. It means that most of temperate country species can't be characterized by barcoding and also all the recent radiations.

So wait that the NGS tools will become cheaper !!!

Cheers from Luc

I already had this discussion with some of my colleagues. I think that I can give some reasons, which, in my opinion are not self-sufficient, or not perfectible, or absolute truthes. And I hope they will not be qualified as offensive to anyone, they represent only my opinion.

For the ease of the reading, I will use T for "Traditional" methods and G for "Genetic"

1) T methods are way cheaper than the G ones, in term of money.

2) Any technician can perform the G's, while the T's require more than the capacity to read a protocol and act according strictly to it.

3) If you don't have the right gene / primer / gizmo, you can identify recent diverged clades (Agree to Luc Legal).

4) The G's use are increasing because of the decision factors / grant providers / other stakeholders - they want to see in the projects / applications / papers "the newest" methods used, while the T's are many times qualified as "old-school", "time-consuming", "obscure" and so on.

5). The T's are used from 1758 on. Their scientific bases have been reinforced in the XXth century, especially after 1950. The G's are used from 1920 or so,, and they have been perfected after 1990. So, the T's have an advance of at least a century compared to the G's. All the sistematics and taxonomy work done until 1990 (and after) are still valid, and over 90% have been confirmed by the G's.

As a G user told me once: I don't need "Classical" T. I need only the names of the species I study. Indeed, the names of the species are useless. And they are given by "geneticists", not by "traditionalists", indeed.

A well-documented paper which supports my opinions is:

Ferdinando Boero (2010), The Study of Species in the Era of Biodiversity: A Tale of Stupidity, Diversity, 2010 (2), pp. 115-126 // doi:10.3390/d2010115 // ISSN 1424-2818 // www.mdpi.com/journal/diversity

Link for the paper: http://www.mdpi.com/1424-2818/2/1/115

Daniel

Funny, as you know I'm subject editor of this journal.....and I reviewed it.

Cheers from Luc

I had already facing same problem that mentioned in your question. I agree with Mr. Cuong Quoc Tang discussions, We can not  rely only on DNA Barcording for Identification of species, we also need a full referenced confirmation of  traditional identification methods for confirmation. It needs for peer reviewed.

Bhavik

Taxonomy is a science, barcoding is a technique used to help taxonomic issues.

I fully agree with Vikas

We should not forget techniques are useful to resolve issues in scientific complications

Hello Luc, I wonder how you know a so exact minimum time of divergence? Three million years also sounds a rather long time to be generalized over life on Earth. Have different subareas of Alps, or Mediterranean islands etc (perhaps not best examples but such that tend to show even quite large, like 7% minimum barcode distances) all really existed so long as isolated? There is sort of evidence from some Lepidoptera, a group of insects, that, say, 10 000 years might sometimes as well be enough to show noticeable differences in barcodes. Tough yes, incomplete lineage sorting etc problems are certainly more frequent between younger taxa - but there is no real reason why they could not persists forever. As mentioned above, there is no molecular clock defining any standard rates of evolution, even between closely related species, not to mention in more distantly related taxa. Nature simply does not work that way. All this said, I do not either use barcodes to nearly anything but screening among "my" very non-descript taxa those specimens that are worth closer study - chiefly I nevertheless use morphology for that, and my taxonomic decisions are always based on morphology. It is of course nice if barcodes match this, but that is only supporting delight, not a defining one.

Hi,

In the future it will be, probably, but at present integrative taxonomy is an essential.

May I advertise a recent paper of me and Adriana Radulovici in which we discussed (beside other thoughts) some examples of problematic relations of barcoding and taxonomy? I would be very interested what you think....

... all the best, Olli

DNA barcodes tell you absolutely nothing about what the organism is (they are just tags like the barcodes of your IKEA catalogue). Barcoding remains time-consuming and expensive (and it does not work when the specimens are inappropriately fixed). The comparison of different barcodes can be used as a useful argument, for estimating whether one or several species are involved, but this is not always conclusive. We have a case where the COI alone does not separate two species (several gene fragments were necessary and there were evident phenotypical differences, in morphology and the colour pattern). At best, in the case of well-demonstrated truly cryptic species (based on the study of several gene fragments), some simplification of the morphological approach might be considered.

A very simple answer is reference sequences. Once we resolve all taxonomical problems and provide reliable reference sequences for all living organisms, we will be able to rely on barcoding. It should take us no more than several hundred years

The molecular and morphological methods can be considered complementary and when applied in combination, constitute a powerful tool for identification with minimal errors. A significant limitation of DNA barcoding alone is the problem of incorrect species identification, which remarkably reduces the reliability and usefulness of the approach. Solutions to reduce these inaccuracies are the documentation of the line drawings and photographs of the species, retention of voucher specimens of the species for later examination.

Because you can use DNA barcoding only on relatively fresh material. Therefore, either we decide that is impossible to recognize fossil (a.k.a dead by more than 1000 yr) specimens, or we need to use also morphology.

As DNA analysis is relatively time-consuming and expensive, morphology and identification keys are essential for field identification during biodiversity estimates and agricultural pest control. Barcoding is highly useful to identify cryptic populations, and to assist taxonomic analysis in deeper research.

I notice several taxonomists have responded, with their frustrations, and rightly so. Genetic barcoding is a valuable complement, and BOLDsystems have been very useful at revealing diversity that was either unsuspected, or in some cases, gone unnoticed, as been reported in few publications already listed here (on RG). This is especially important when comparing global data, which is now easily visible under BOLDsystems, as taxonomists and biologists may have worked for decades with regional specimens, without comparing wth others. However, barcoding has expanded very quickly, and as Charles Oliver Coleman and Radulovici point out in their article, the metadata, including identifications and photos, are sometimes flawed or in error. The current approach, similar to that of the citizen science platform, iNaturalist, is to allow records to be posted with at least some information, and hope for corrections (in the case of BOLD, the barcodes are grouped together based on their sequences, so mistaken identifications can be noticed; with INat, users reply based on photos to correct the identifications). Unfortunately, among the many interesting barcode submissions are a good number of flawed records that are not always so easy to correct, especially when they get published in review articles. Cue the disappointment. My answer is barcoding is a tool, that can be used (and abused). Easier updates/corrections and oversight will be needed to make best use of them (similar to the situation with occurrence records on the biodiversity platforms GBIF and OBIS).

Simply, genes do not express everything into morphological traits always, even though genotypes give many clear answers scientifically. Therefore, genotypes could not be seen the exterior (surface) of organisms. Consequently, we need to have knowledges of phenotypes in order to understand organism, if the phenotypes correlate with genotypes.

I think it is a good idea to use the morphology, just for confirmation. By the way, for "Species" identification, only one pair of primers will not be enough.

Because species cannot be confidently identified using DNA barcoding, not least because of the many misidentifications already in the system and the lack of genetic difference in many sibling species that can be separated morphologically and/or ecologically.