Tuesday, August 28, 2012

A Facebook Conversation

  • Below is a conversation on Facebook started by my colleague Frank Pezold in response to my first posting on the Fish Phylogenetics blog.  For those who may not know, there is a group of ichthyologists that are not pleased with the use of data matrices and optimization methods for the inference of phylogenetic relationships. Really, this is not a joke.

    These systematists are particularly displeased with the use of molecular data in these efforts and are concerned that "molecular classifications" will make systematic ichthyology irrelevant.

    While the concern is pointed at the ichthyological taxonomic discipline, it actually represents a fringe element of the phylogenetics community that continues to advocate three taxon analysis. Also, from a wholly misinformed perspective, they refer to data matrix driven phylogenetics as "phenetic." 

    Let me note that this perspective has absolutely no traction in the mainstream field of phylogenetics. These papers are not being published in Systematic Biology, and it is unlikely they ever find a larger audience. My guess is that this very narrow view of phylogeny is maintained in ichthyology because we do not have a well-resolved set of phylogenetic hypotheses that includes the diversity of all living teleost lineages. Once such a set of phylogenies are available, I suspect these views will retreat from the landscape of practicing phylogenetic ichthyology. It is important to note, as is obvious in reading these comments, that a resolved phylogeny of teleost fishes is not a goal of these systematists.

    I think that we as phylogenetic ichthyologists will continue to remain irrelevant if we are not able to articulate the relationships of teleost fishes to the broader community of scientists attempting to resolve the entire Tree of LIfe.

    The conversation is posted just as it appeared on Facebook.

    Frank Pezold Nice site Thomas. I have a comment on this comment - "Much of what has constrained the development of ray-finned fish phylogenetics is been the reluctance of many ichthyologists to whole heartedly incorporate molecular data (e.g. DNA sequence data) in their efforts to infer phylogenetic relationships." I think the difficulty primarily arises from speedy conversions of molecular phylogenies to classifications, which can be fraught with problems much discussed, and the reduction of information available in morphological studies to a simple overlay on a molecular phylogeny instead of its use as another dataset. neither of the problems is attached to this publication...it offers an independent analysis that has resulted in significant phylogenetic insight.
    August 24 at 12:53pm ·  · 3

  • Thomas Near Frank, thank you for the comment. You should post this on the blog, where I can provide a longer response. What I will say here is that I rather see speedy conversions of molecular phylogenies to classifications than the continued recognition of para- and polyphyletic groups in the established classification.
    August 24 at 1:06pm ·  · 1

  • Frank Pezold there is a middle ground i think that doesnt require the adoption of the latest hasty and often incomplete analysis yet doesnt rest forever on the rump of a stodgy ignorance of the value of molecular tools. I also fear that much of the housekeeping left behind the wash of the molecular systematic wave will be left to us older doffs or just left behind period. i'll check in on the blog, but for the moment this has more readers, and it is on my radar. glad to see all of this - the paper and discussions online.

  • Thomas Near Can you cite some examples?

  • Frank Pezold Of what, the fact that the two ends of an extreme are joined by a continuum?
    Saturday at 9:34pm via mobile · 

  • Thomas Near No, hasty changes to classifications.

  • Frank Pezold Of proposed classifications, sure, but I thought the question here was of philosophical approach.
    Saturday at 9:39pm via mobile · 

  • Anthony Gill ‎"...I rather see speedy conversions of molecular phylogenies to classifications than the continued recognition of para- and polyphyletic groups in the established classification." This implies that the molecular methods are somehow retrieving truth ... or perhaps that morphological studies are not questioning "established classification." Neither is true. The essence of systematics is understanding characters (homologues) and their distributions ... something that is not generally tracked in molecular studies (only phylogenies and statistics are presented). Rapid changes in classification from studies (morphological or molecular) that do not make a serious attempt to understand characters are unlikely to be a good thing. More comments on this (and the fate of molecular classifications) is given here:http://www.mapress.com/zootaxa/2010/f/zt02450p040.pdf
    Sunday at 10:13pm ·  · 1

  • Thomas Near I said neither. That is what you want to see. No one who practices phylogenetics and realizes that the science is based on inference thinks they are revealing "truth" when they present any phylogenetic hypothesis. The whole meme that molecular phylogenies do not track the character distributions is at best tired, and at worst disingenuous. One can take the phylogenetic tree that represents an optimal distribution of character stats and the data (phenotype or DNA sequences) to assess character support. There is not enough printed space in a typical journal article to map the character changes. Also, any statement that these phylogenies are not the result of optimal character state distributions is uninformed. Yes tress and support are presented...in every phylogenetic field in the early 21st Century. We can pine for the glorious days of gas-lit streets, but those days are past. So are noteworthy phylogenetic studies that sample 8 taxa for 30 characters. As for classification, I would still prefer the "rapid changes" to the clearly wrong aspects of teleost classification that everyone knows are not supported by available data (e.g., Protacanthopterygii and Perciformes), but are maintained because of some unspoken standard that is required to change the revered tablets handed down by Regan, or from the shear force of personality for particular pet hypotheses (e.g., Johnson and Smegmamorpha). Ichthyology is in a crisis, not because of molecular inferred phylogenies and rapidly changing classifications, but rather our crisis is tied in our inability to articulate our work in the context of 21st Century phylogenetics. We have the most difficult problems in the Vertebrate Tree of Life, but we fail to capitalize on these interesting questions to get systematists in other organismal disciplines interested in our work. Instead of using the best tools available, we squabble while the largest polytomy (Percomorpha) in the Ray-Fin Tree of Life remains unresolved.

  • Anthony Gill Yes, inference it is. But I see words like "resolved" thrown around in recent molecular papers, which suggests that truth has been revealed. I see the current emphasis on phylogenetic trees (solutions to data sets) and statistics as at odds with character driven investigation. Current molecular phylogenetics (and other optimization approaches) do not simply track character distributions, they create them. Character conflicts in the original observations are reinterpretted as new characters and evidence for new relationships. I do not see an obvious compromise, because it seems that the different approaches have different goals (which is not to say that they won't sometimes deliver the same relationships). One is an investigation of character evolution (and driven by particular models of evolution), while the other is an investigation of character distribution and classification (a pattern approach). Perhaps I am from the days of gas lights (though the light source for my microscope is LED), but I often find more worth in an overlooked paper by Theodore Gill than in the latest phylogeny and classification of x or y, because Gill described characters that could be reinvestigated. Gill's contributions in the long run was not his classifications or his suggested relationships, but the evidence that formed the basis of his classifications. This is not to say that our existing classification system is perfect and there is no shortage of dogma (I commented on the Paracanthopterygii and Perciformes problems nearly 20 years ago), but simply creating a plethora of short-lived new ones is likely to make our field irrelevant.
    Yesterday at 12:23am ·  · 1

  • Thomas Near No words are being thrown around, they are chosen carefully. This is reflective of the fact that some of these papers are coming out in the most difficult journals to have one's work published. I see the word resolution as relating to the ability to strongly discriminate among alternative answers given a particular model and set of data. It is your suggestion that resolution means truth, not mine nor is it the unambiguous definition of the word in the OED. If you are at odds with the current state of affairs in phylogenetics, then your target needs to be much more inclusive than ichthyology. Even when ichthyology catches up with the other vertebrate disciplines in the sophistication of its phylogenetic work, we as ichthyologist will still maintain our connection to previous workers through our excellent approach to scholarship. Yes, even a molecular systematist can appreciate your overlooked paper by Gill and the character information it contains. I would also posit that molecular characters can, and often are, reinvestigated. Molecular data is very fine evidence for phylogenies, and hence any classifications that would follow from those phylogenetic perspectives. There appears to be an odd perception that molecular data is not evidence. It is. Even if a worker does not map their characters on the trees, it is still evidence. I appreciate your comments on groups like Paracanthopterygii and Perciformes; however, there they are, still ensconced in some of the most referenced teleost classifications. Let's stop beating around the bush. Tell me if morphological data is ever going to result in a resolved phylogeny for let's say 500 percomorph species sampled from more than 200 taxonomic families. If this is possible, when will it happen?

  • Anthony Gill As I said, different goals.

  • Thomas Near So a goal of yours as a systematist is not resolution of the Tree of Life?

  • Prosanta Chakrabarty Tom, please cut and paste this discussion on the new blog, so that the discussion is preserved - hopefully it can continue there
    23 hours ago · 

  • Prosanta Chakrabarty Anthony GillFrank Pezold, please like this or the above comment to give Thomas Near permission to repost this discussion to the phylogenetics blog.
    23 hours ago ·  · 3

  • Luiz Rocha Yes, would be great to have this discussion preserved and extended...
    22 hours ago ·  · 2

  • Rodrigo Torres Niiice!
    22 hours ago · 

  • Anthony Gill I see a large number of goals in your post Tom, many of which are sociological rather than scientific. My goal as a systematist is as it has always been: to continue to develop a natural classification of fishes through improved understanding of characters and their distribution. My understanding of character distribution is based on a meagre collection of around 800 species (about 2,500 specimens) representing about 150 acanthomorph families, but with reference to a much broader literature-based survey across fishes - supplemented with about 30 years of direct observation of specimens. My goal is to diagnose groups - natural classifications - not ancestral conditions at nodes. That, in a nutshell, is the fundamental difference in our goals. I see no particular need to repost my comments elsewhere, or really to devote more time to any discussion.
    17 hours ago · 

  • Thomas Near No sociology here. Just wanting to know how teleosts are related and their divergence times. If you truly have a 800 species dataset, you should publish that in a comprehensive paper. It would be a wonderful and impactful contribution to the field of systematic ichthyology. You may retort that you published that data in various papers, but it would be nice if you had an ambition for a grand synthetic work. This is reflected by the fact that I was not aware of your incredible dataset because it has never, to my knowledge, been published in one piece of work. By the way, monophyletic groups are the basis of natural classifications for 99% of practicing systematists. I guess you are are a "one percenter." Please excuse my light-hearted joke. It is meant to reflect that fact that all of us ichthyologists are in the same boat.
    16 hours ago · 

  • Anthony Gill It is obvious again that we are not on the same page. There is no dataset as such, because that is not how morphological systematics has progressed. The mention of the specimens (which only refers to skeletal preps on my lab bench - there is a similar set available for study at ASU, and I also based many observations on skeletal preps in the NHM, AMNH and USNM) is simply to contrast how we contribute (and that we do not indeed have the same goals). You challenged morphologist to produce a topology based on 500 species across 200 percomorph families, because that is the only way molecular studies can contribute (at least while attention is solely on topologies rather than characters): each topology is a solution to a given dataset, which stands in isolation from future or past solutions. In their quest for understanding characters and their distribution, morphologists routinely refer to much larger "datasets" - specimens (the only realities we actually deal with) and literature (which may guide us to examine other specimens). I could easily amass a huge number of characters that spanned specimens available to me for study, but many of those characters would be poorly defined and not representative. (I could easily populate it a matrix with fin-ray and scale counts, for example.) The emphasis, instead, is on trying to clearly define characters and understand their distribution. It is not merely about large numbers of characters - is the weberian apparatus not adequate to diagnose a group of fishes? You are correct in putting me in a 1% minority, however, in that I am convinced the current emphasis on optimization methods is problematic. It is not that I do not recognise natural groups as monophyletic, but for me monophyly indicates that two or more taxa are more closely related to each other than another taxon outside the monophyletic group. Refences to processes and ancestors (or ancestral characters) lie outside our field of observation, and may actually impede rather than assist in developing an understanding of relationships. I acknowledge also that I am unlikely to make a grand contribution to the tree of life. I am in a non-research position doing research in my spare time supported out of my own pocket. However, I hope I can make a lasting contribution to what I see is the essence of systematics - the understanding of characters and their distribution. It is these and not ever-changing topologies that matter to me. I'm afraid that same limiting resource (spare time) is why I cannot continue to dedicate time to this discussion, or to pursue it on other portals.
    14 hours ago ·  · 1

  • Prosanta Chakrabarty Tony, you don't need to continue the discussion if you don't want to, but please grant permission to post this on the original blog so that a broader group can see it and it can be saved in a better place.
    13 hours ago via mobile · 

  • Anthony Gill Okay, go ahead.
    13 hours ago · 

  • Thomas Near I think it is funny that you are turned off by "ever changing topologies" when that is the essence of inferential science, changing hypotheses with the addition of data. Given that the node which subtends nearly 17,000 percomorph species is currently a polytomy, we are facing a period of flux. Going from a polytomy to any resolved solution is a change. I am confident that the ambitions of us who hope to bring systematic ichthyology into the 21st Century will be realized. Please continue to contribute your views, as all voices in our science are welcomed and appreciated.
    13 hours ago · 

  • Frank Pezold This has been very interesting but I think we have drifted a little from the spark that lit the conversation. It has to do with the responsibility we have to users of our information (those outside of the circle of systematics). Working from morphological studies systematists would amass data, as Tony describes, and periodically publish a tome sometimes in concert with data collected by others. This usually took significant time and much was vetted in talks prior to the publication. Molecular data is much more quickly gathered and processed and this has led to rapid publication of new hypotheses - whether of species or phylogenetic relationships. A major concern I feel is that we don't have the same period of vetting and whether it is a question of what constitutes a species or a higher taxon, we have an often fluid vision of relationships (dependent on the particular gene or the limits of taxon sampling) that can change quickly. It is not a question of whether or not one dataset is better than another - we need all the data we can get while of course weeding out irrelevant data from either set. We just need to show some sense of responsibility when we share our results. Otherwise we trivialize our work and cheapen the role of the systematist in evolutionary biology. I found the work shared on this post informative and look forward to more from the authors as well as insights we will obtain from the information Tony is gathering.

No comments:

Post a Comment