I don't see a problem with publishing single-gene phylogenies, so long as the limitations of such analyses are clearly stated. I think that, yes, any answer is better than no answer and this is the basis of scientific progress. All studies are limited (even a phylogeny based on a whole genome might not be truly representative of a whole population/species) and represent our best knowledge to date. We may still be wrong but we should be getting progressively less wrong. If someone is trying to publish a single (or limited) gene phylogeny as THE definitive relationship, however, that is clearly bad science and they should know better. This is particularly true when horizontal gene transfer and/or hybridisation is known to be common.

I don't see a problem with publishing single-gene phylogenies, so long as the limitations of such analyses are clearly stated. I think that, yes, any answer is better than no answer and this is the basis of scientific progress. All studies are limited (even a phylogeny based on a whole genome might not be truly representative of a whole population/species) and represent our best knowledge to date. We may still be wrong but we should be getting progressively less wrong. If someone is trying to publish a single (or limited) gene phylogeny as THE definitive relationship, however, that is clearly bad science and they should know better. This is particularly true when horizontal gene transfer and/or hybridisation is known to be common.

If you pursue puritanical conceptual approach, and considering conflicting gene tree topologies, the term "phylogeny" would apply only to non-recombining loci and not to organisms (and moreover populations). Therefore, any "organismal tree" or "species tree" is rather an approximation that simplifies to present species evolution as a tree rather than as a network. This is especially true if lower-level taxonomy is considered. For this reason I don't find fruitful to discuss how many loci one should consider for constructing species-level phylogeny. It were much more correct to discuss gene trees, their coincidence or discordance if they belong to the same organism and just forget "true species tree" as a fiction.

If no data is available for the group of interest, a phylogenetic tree reconstructed from only a single gene seems to be an interesting first step. Then, phylogenetic hypotheses can be tested with additional markers.

My experience with plants is limited, but I know that the chloroplastic genome evolves more much faster than nuclear and mitochondrial genomes.

So, if your phylogeny involves closely related species, you should first focus on one chloroplastic marker, and then select a series of nuclear markers... just in case of past events of interspecific hybridization.

Very simply: being adamant about it and makig very clear that what we are publishing is a gene phylogeny, not an organismal phylogeny. That different parts of the genome have different eviolutionary histories is a reality that should be familiar to any reader of a scientific article in a journal of evolutionary biology. There is nothing wrong in publishing single-gene or plastid gene phylogenies, what is wrong is interpreting them erroneously.

Nowadays I see a lot of early-career researchers sitting on single-gene mtDNA datasets for fear they would not pass peer review. I would say that if you have a plastid or mtDNA dataset, for goodness sakes publish it. You'd have to take the conclusions with a grain of salt obviously, and as Richard states above, make sure you describe the limitations of single-gene studies and any potential avenues of future research. It represents the branching order of one gene only but still certainly constitutes valuable knowledge about the evolutionary history of the group of interest. How many well-established folks in the field built their careers on these kinds of papers? The answer is: a lot. Today, people feel very strongly about accounting for coalescent stochasticity, which is important, but not everyone has the budget to produce many gene phylogenies much less NGS datasets. However, the base knowledge of cpDNA or mtDNA can provide testable hypotheses with NGS datasets if you were to acquire them in the future, as well as preliminary data for a grant that would provide the funding for a larger analysis.

You may have read these:

Zink RM, Barrowclough GF. Mitochondrial DNA under siege in avian

phylogeography. Mol Ecol. 2008 May;17(9):2107-21. doi:

10.1111/j.1365-294X.2008.03737.x.

Edwards SV. Is a new and general theory of molecular systematics emerging?

Evolution. 2009 Jan;63(1):1-19. doi: 10.1111/j.1558-5646.2008.00549.x.

There is an extensive literature on this question. Many of the papers you cited have been critiqued by others, for example

Kluge, A.G. 1989. A concern for evidence and a phylogenetic hypothesis of relationships among Epicrates (Boidae, Serpentes). Syst. Zool. 38, 7-25.

Brower, A.V.Z., DeSalle, R.,Vogler, A.P. 1996. Gene trees, species trees and systematics: a cladistic perspective. Annu. Rev. Ecol. Syst. 27, 423-450.

Gatesy, J., Desalle, R.,Wahlberg, N. 2007. How many genes should a systematist sample? Conflicting insights from a phylogenomic matrix characterized by replicated incongruence. Systematic Biology 56, 355-363.

Basically, if you are not willing to accept that the hierarchy of life is empirically accessible via evidence from characters, then you probably should not be a systematist.

Basically, one shouldn't mix up phlogenetic studies with taxonomic ones. I use to build tonns of phylogenetic trees, all of them are describing a history of a particular gene/gene family, not pretending to describe the history of the species. So far, I did not encounter any misunderstanding or misinterpretation of such trees.

The idea of the individual history of each gene is optimally reflected by the consideration of phylomes (published by Toni Gabaldon in 2007), when each species is characterized by the set of phylogenetic trees built for each individual gene. Clearly, these trees do not coincide within one species. You can use 100-gene or so-called "all-genome"-based phylogeny, - it will not correctly reflect the history of each and every gene, there always will be exceptions. You just have to know about it and interpret the data correctly. Here I completely agree with Fabio Rindi.

There are some really good answers and tons of information contained in the responses posted so far. But just to add my 2 cents worth, the single gene discussion should also include consideration of the so called “bar-coding” movement which has received much funding as well as much criticism. One position criticizes the use of a single mtDNA marker because the marker provides misleading patterns of variation; specifically, phylogenies that are inconsistent with those derived from nuclear gene sequences in the context of species relationships among closely related taxa. On the other extreme, the DNA barcode movement promotes the sole use of small fragments of a single mtDNA gene, cytochrome c oxidase I (COI) to identify most of life (see Hebert et al., 2003). A number of researchers including myself and Dan Rubinoff several years back, joined the fray (Rubinoff and Holland 2005, Systematic Biology) pointing out disadvantages of both extreme viewpoints!

We, and others, have argued for an integrated role for mtDNA and single gene phylogenies, one that takes advantage of mtDNA’s many strengths but also accounts for its shortcomings by using it in concert with other independent data sources (e.g., cytosystematic, morphological, behavioral). We are against the abolition of the use of single gene phylogenetic reconstruction, particularly as a starting point, as other ResearchGate posts have suggested, as this approach clearly has biological value in many fields including phylogeography and conservation (Holland and Hadfield 2002, Molecular Ecology).

But I would also argue against use of a single gene as a “final answer” to understanding any lineage history, and narrow use in barcoding as currently defined. Again to my way of thinking, neither viewpoint is particularly productive and there is no doubt that single gene phylogenies can be an important, timely, cost-effective tool in the context of initial taxonomic and phylogenetic studies.

The contentions on the one hand that mtDNA is misleading and on the other hand that it is universally applicable for systematic identification and reconstruction of evolutionary history are not mutually tenable. Neither viewpoint considers the appropriate role that mtDNA can play in systematic studies because both fail to consider the information provided by the gene in context.

I agree with the publication of single, two, three etc gene trees for a group of organisms, especially if there are not data out there on a hypothesis of phylogenetic relationships. Too many people and way too many scientists think that what ever they publish is the answer to the question and if someone comes along and offers an alternative resolution, then the latest published and the one with more genes must be correct. However, they may all be wrong for a variety of reasons that most of you likely know about. So, the thing is there is an abundance of diversity for which we have NO information. Thus, let the gene tree/species tree thing go for a few years and let this all get worked out in the literature. Holding back information and leaving single gene tree papers in a file folder does nothing to advance science. With the sequences deposited in public places then people can collaborate in sharing genes and continue to work on relationships of organisms. At the same time, those people working on algorithms that involve lateral gene transfer or lineage sorting problems have a treasure chest of data to work with. Science has to move on and holding people or papers back due to some artificial number of genes or where the genes come from (mt, N) is not science. In my opinion it is simply a close-minded view of the world and can be a lack of peripheral vision as to what science involves and the uncertainties that we all face day in and day out. Cheers.

Every method has its field of applicability, phylogenies (single-gene included) not being an exception. One just has to keep this in mind and use them correctly, applying to feasible problems. It would be unfair to put in question a whole huge field of research, which has a large impact on our understanding of particular part of biology, only because of the risk of inadequate application of the method and interpretation of results.

One issue I haven't seen addressed above in an admittedly cursory skim of the other answers is the nature of the group of organisms you're considering, and the associated degree of divergence among them. A group of multicellular organisms as recent as the great apes, for example, is extremely unlikely to be subject to lateral gene transfer. However, a speciation "event" may involve isolation of one subpopulation of apes from another, and therefore more than one allele of a marker may pass through the "event" into both nascent species, and these markers may persist over a small number of generations to the present. Thus, sequencing one individual from each species may show any of the four marker patterns (AA, Aa, aA, or aa), purely depending on which individual was sequenced. These markers can be as large as synteny blocks, and thus may contain entire genes.

The above issue doesn't need to be considered when discussing the phylogeny of the plants, because neutral variation is extremely likely to fix on one or the other variant by genetic drift over 1 billion years or so of evolution. Instead, the concern should be whether there was one plastid endosymbiosis event, or multiple, during the evolution of the plants. I believe the likelihood of endosymbiosis into the germline is seriously reduced for multicellular organisms; so whether this issue is relevant depends to some extent on how large a group of plants you consider. Thus, a gene tree built from a single plastid gene MIGHT be relevant for a clade of plants believed to have been multicellular throughout their evolution. However, since gene transfer between the nucleus and organelles has been observed in both directions, one must also consider whether the plastid gene has been carried in the plastid genome throughout, or whether it has shuttled between the plastid and nuclear genomes.

Unicellular organisms, whether protists, plants, animals, bacteria, or archaea, are more susceptible to lateral gene transfer than multicellular organisms, and in fact unicellular organisms appear to have active mechanisms for importing entire operons. In the case of phylogenies including unicellular organisms, especially over long evolutionary distances, single-gene phylogenies should not be trusted.

Thanks for these replies. It seems that the underlying belief is that as long as a gene region can be aligned and homology can be assessed this is going to allow us to capture important evidence about the evolutionary history of the organisms sampled. Keeping in mind that this is just a genealogy (from a gene) we still think that this will mostly point toward "real" or "true" evolutionary relationships of the organisms we sampled. I really hesitated to use the word "real", or "true", I know it is just a hypothesis but to me that is the crux of it. In the end we are only producing a model of reality. We are justified in taking this approach as long we acknowledge that evidence may mislead about some of the relationships for all the typical reasons, its more complex that we can easily comprehend , extinctions, sampling artifacts, lateral gene transfer, duplication, parallelisms etc. Do you agree that a "species tree" summarizes evidence collected from more loci and or other characters? And that allows us to artificially delimit the biodiversity we study in meaningful ways? It seems we are often interested in species trees because these allow us to organize our knowledge about the world and communicate about it.

Allow me a small correction: this is going to allow us to capture important evidence about the evolutionary history of the GENE sampled. Whether this reflects or not the evolutionary history of the organism sampled, it is our job to understand it. After all, we are biologists; besides being proficient with sequencing and phylogenetic analyses, it is fundamental that we understand in depth the biology of the organisms we work with.

I agree with you that species trees allow to organize knowledge about the world. The point is: what is really a species tree? How many markers or characters should be included in a dataset so that the tree inferred can be considered a species tree? I am afraid things here become very subjective.

This question has a subtext that wonders whether more is better when it comes to the number of genes that contribute to phylogenetic analysis. One expectation might be that a wrong answer from a single gene would be corrected if many genes are added to the analysis. In fact, there are cases where the error is made worse by adding more DNA data, suggesting that more attention should be made to specific corrections to DNA data rather than brute force solutions that merely stack on more data. 

My favorite example is the eucaryote slime mould, Dictyostelium, whose DNA, in huge amounts, including a whole genome, "says" that it is a procaryote! This error is the result of its massive A+T content, at 85% the highest of any eucaryote, which leads to coincidental A or T substitutions that are shared with some procaryotes that also have high A+T content, a kind of chance molecular convergence that has occurred without the usual selective evolution. If the phylogenetic reconstruction uses amino acid sequence, instead of the encoding DNA sequence, Dictyostelium moves a number of nodes up the tree of life to join other eucaryotes, like the fungi, where it belongs. This simple test, to run separate phylogenies using both DNA and translated codons from the same sequence, should be used more often when there is dissension caused by the possibility of convergent DNA. 

I  have repeatedly advocated the use of this test in the controversial case of the phylogeny of the two kinds of bats. I will not go into all the details of this case, except to point out that virtually all of the characters used to support the monophyly of microbats and megabats are flight-related and so may have arisen independently in two unrelated clades of flying mammals whose separate origins are suggested by other evidence, such as brain and amino acid sequence. In other words, the test for convergent DNA is positive for bats, given the completely different phylogenies obtained from DNA (monophyly) and amino acid sequence (flying primates). A lot of emotion surrounds this case, partly because some wrongly think it represents an attack on the whole edifice of DNA phylogeny instead of an earnest attempt to introduce corrections for DNA in the rare cases where it can lead us astray. In support of this constructive general view of DNA phylogenetic reconstruction,  I can cite my own recent use of a single gene (ITS) phylogeny to describe a new species of baobab, which is a propos the present question by showing that a single gene can be adequate in many cases. On the question whether a positive test for DNA convergence was "on the money",  Recent whole genome studies of bats confirm the presence wide-scale convergence of DNA between megabats and microbats in their immune systems. Each immune system is massively enlarged as a result of flight-related increase in metabolism, but via two completely independent genetic mechanisms. 

Convergent DNA is not a common event, but there are tests for it like the one above. Piling on many extra genes will not help the precision of the phylogenetic reconstruction if the test is positive and uncorrected.

There are also molecular markers that are not specific to a gene that can be useful in constructing phylogenies; see for example Suh et al. 2011 (http://www.nature.com/ncomms/journal/v2/n8/abs/ncomms1448.html). One of the values of these non-genic markers is that they are expected to be selectively neutral, and therefore are unlikely to occur at corresponding positions in separate lineages.

To all

I think it is just a story of semantic.

Now building tree(s) with a single gene is namely "barcoding" (see all the studies using the single gene Cox1). It gives an image of relationships between sampled species but rarely a relevant Phylogeny.

End of the 90' to propose phylogenies with 2 genes was a high standard work (publications in Mol Ecol or Mol Phyl Evol).

Now it is difficult to name a study: "phylogeny" if not using 5 to 7 genes.

Anyway studies with a single gene are always giving information and may lead us to determine for which type of organism it is valuable to perform a "real" phylogenetic study.

My proposal is to be clearer of what we are naming "barcoding" and what is a phylogenetic study...knowing that it is constantly changes,....more it will go more we will have access to larger part of the genome of our samples.

Dear Chris and all,

A central, yet often forgotten point in phylogenetics inference is that this is a fundamentally statistical process. We just can't pretend to know (exactly, precisely) the Tree of Life --pretending otherwise is self-delusion, and an unfathomable source of misunderstanding.

So, a one-gene tree may be a very good option depending on what your question is, and/or how robust that tree is. Think of your quest as that of an archaeologist tryng to put together an epic poem written in cuneiform alphabet on broken clay tablets. A computer obviously helps, but the problem remains the same: you gain confidence on your interpretation as more data confirm previous steps, or fill in gaps reasonably well, thus rendering a better picture of the whole story. It often happens that such a text can be recomposed readily (one gene). In some cases, it will be very hard, requiring a variety of approaches (say, many genes).

Now, the fact that nucleic acids have only four possible character states for each position implies that telling apart phylogenetic signal from noise is no trivial game. We talk about homologous sequences because they are statistically so similar that chance alone would hardly yield such data. We trust the trees produced by a computer program because we understand the indices showing statistical support. However, the four-state limitation leads to artifacts --more often than usually acknowledged. Long-branch attraction, for example, is so vicious that a wrong answer will gain support as more and more data are added. An important point when we see the tsunami of long sequences coming.

Thus, in my opinion the answer boils down to this: it's not quantity, but quality. The good news is that size doesn't matter. The bad news is that you have to know what you are doing --and just come to terms with your limitation.

All the best,

Cristian

With a rapidly increasing number of whole genome datasets being analysed in an evolutionary or phylogenetic framework, the inconsistencies in coalescence of genes within a single genome will become ever more apparent. This will highlight the evolutionary significance of genetic introgression due to hybridisation and horizontal gene transfer, and I suspect it will show that such genetic exchange across taxonomic groups could be an important source of novel variation that fuels adaptive evolution. It may also shed some light on other evolutionary phenomena such as trans-species polymorphism.

Recently, we analysed five genomes of an oomycete plant pathogen which showed that the phylogenetic relationships differed depending on which genes you analysed. Well supported gene trees of the same five OTUs (i.e. in our case five pathogen races) looked completely different depending on the set of genes that were being analysed. We suspected this was caused by genetic exchange (introgression) between the races and we designed software to visualise this signal across the genome. The genomes had a mosaic-like structure and we uncovered hundreds of introgressed regions which coalesced at variable amounts of time (between ~10^3 and 10^5 years). It shows that the inference you can make by analysing the phylogeny based on a single gene has obvious limitations especially when the evolutionary signal is muddled up by introgression. However, when you aim to test, for example, a phylogeographic hypothesis, mitochondrial genes might currently still be the most feasible genes to analyse.

The software is called HybRIDS (Hybridisation, Recombination, Introgression & Dating Software) and can be freely downloaded from here: http://www.norwichresearchpark.com/HybRIDS

Can HybRIDS be used with hundreds of loci and hundreds of individuals?

HybRIDS can be used to analyse millions of SNPs, but in the current version the number of individuals is limited to tens of individuals.