Tuesday, June 30, 2009

Genes, Phylogeny, and Orangutans

Jeffrey H. Schwartz is well known to talk.origins veterans because we discussed his book (Sudden Origins: Fossils, Genes, and the Emergence of Species) back in 1999. Schwartz tried to make the case for a "groundbreaking and radical new theory of evolution." This "theory" was based on the idea that new species spring into existence very quickly when a mutation in a homeobox (HOZ) gene arises in a population. It's a "theory" of saltation but it's based on such a flawed understanding of genetics that you really have to read to book to see just how bad it is. Sudden Origins is a leading candidate for the worst science book ever published.

In case you want to see a shorter version, the basic idea is explained in Schwartz (1999).

Over the years, Schwartz has published many other ideas that are controversial. Lately he has been pushing the concept that molecular phylogenies are unreliable. In part this is because he is opposed to gradual change as documented in the record of the genes. He thinks that real evolution takes place when alterations of regulatory genes result in major new phenotypes. Thus, the best way to discover the history of life is to examine anatomical homologies and differences.

But part of the problem lies in Jeffrey Schwartz's idiosyncratic understanding of genetics and molecular biology. When you put these together, this is what you get in Schwartz (2009).
This having been said, systematics and evolutionary biology need not remain estranged. Developmental biology increasingly makes clear that organismal change (and by extension, evolution), is not how it was imagined when the synthesis emerged (see reviews in Schwartz 1999, 2009b; Maresca and Schwartz 2006). Further, because of the interrelation between, e.g., the physical properties of cells, signaling pathways, epigenetic effects and development and consequently the origination of form, the false dichotomy of ‘‘molecules versus morphology’’ that resulted in the 1980s from the dethroning of morphology by the hegemony of molecular analyses is no longer tenable (Schwartz 2009a). Indeed, the undeniable hierarchical continuum from the molecular through the morphological, firmly centralizes morphology (as understood via development) in systematic endeavors (Schwartz 2009a).
Grahan and Schwartz (2009) have just published a paper in which they claim that orangutans are more closely related to humans that are chimpanzees. According to them, the molecular data is not reliable. They claim that detailed morphological comparisons show that orangutans are our closest ancestor.

John Hawks asks the question "Are orangutans our closest living relatives?" and he comes up with the best possible answer to scientists with a well-known history of promoting "unusual" positions on evolution.

It's a lesson that New Scientist should have learned. They devote several pages to the Grahan and Schwartz paper thereby giving it much more publicity than it deserves [Could the orang-utan be our closest relative?].1 The article is written by Graham Lawton who you might remember from the "Tree of Life" episode [see: Explaining the New Scientist Cover]. The editors of New Scientist knew full well that their decision would be controversial so they took a proactive position by writing a short editorial [In praise of scientific heresy ].
If its claims are so outlandish, should the research even have been published? Some scientists would clearly have preferred it if the paper had never seen the light of day, and question the judgement of the journal.

That is territory we should tread with care. Ideas that mainstream opinion "knows" to be wrong occasionally turn out to be right. The insights of Galileo, Stan Prusiner - who discovered prions - and many others were once denounced as heresy. And even those that are wrong can be valuable.

Science proceeds by questioning its own assumptions and regarding every "fact" as provisional, so alternative hypotheses should be given an airing, if only to reaffirm the strength of the orthodoxy. Science that pulls up the drawbridge on new ideas risks becoming sterile. The journal recognised that and should be applauded for its decision to disseminate this challenging paper.
There's some truth here, but only some. You can't use Stanley Prusiner as an excuse to publish every crazy idea that comes along. Some ideas really are crazy—they are not revolutions in disguise. The plain fact is that Jeffrey H. Schwartz has already had his chance to make his case and he has not been successful. How many chances does he get before we draw the obvious conclusion?


1. The article was two pages long and the editorial was much less than one page. This may not qualify as "several" pages by some definitions.

[Photo Credit: Daily Mail]

Grehan, J.R. and Schwartz, J.H. (2009) Evolution of the second orangutan: phylogeny and biogeography of hominid origins. Journal of Biogeography, published online June 22. 2009. [doi:10.1111/j.1365-2699.2009.02141.x]

Schwartz, J.H. (1999) Homeobox genes, fossils, and the origin of species. Anat Rec. 15:15-31. [PubMed]

Schwartz, J.H. (2009) Reflections on Systematics and Phylogenetic
Reconstruction. Acta Biotheor 57:295–305 [doi: 10.1007/s10441-009-9078-9]

27 comments:

  1. I don't haveaccess to the paper, but I wonder if he cites this paper:

    Shoshani, J., C. P. Groves, E. L. Simons, and G. F. Gunnell. 1996. Primate phylogeny: Morphological vs molecular results. Molecular Phylogenetics and Evolution 5:102-154.

    They used over 100 anatomical characters and found that human linked more closely to chimp than either gorilla or orang.

    ReplyDelete
  2. I don't haveaccess to the paper

    Although the journal is not open access, the editors posted a PDF at:

    http://www.blackwellpublishing.com/pdf/jbi_2141.pdf

    ReplyDelete
  3. LArry,

    Grehan has shown up over at my place defending the paper.

    ReplyDelete
  4. Funnily enough, Grehan is at the same institution as Denis McCarthy, one of the remaining proponents of the expanding earth (although McCarthy left recently I think). Grehan is also a proponent of panbiogeography, which is a pretty controversial position I think (and funnily enough, the main panbiogeography page on the net contains a link to expanding earth stuff, along with conventional geology). There's obviously some heterodoxy in the water up at the Buffalo Museum of Science!

    ReplyDelete
  5. This comment has been removed by the author.

    ReplyDelete
  6. Perhaps it's a generational thing - I've been immersed in molecular methods since the start - but I simply can't fathom how the people arguing that molecular phlyogenetics is a load of hogwallop can make their argument if they've ever heard of the phrase "homoplasy..."

    ReplyDelete
  7. @ Doppleganger

    They cite the paper only to note that it omits characters that would unite Homo with Pongo.

    ReplyDelete
  8. The plain fact is that Jeffrey H. Schwartz has already had his chance to make his case and he has not been successful. How many chances does he get before we draw the obvious conclusion?

    You ain't seen nothing. Papers claiming that birds are not dinosaurs are still being published.

    ReplyDelete
  9. RE the remarks by New Scientist:
    Unconvential ideas should not be blocked for being unconventional. They should be blocked if they fail to meet standards of evidence and argumentation. I.e. the bar should not be lowered for them.

    ReplyDelete
  10. Happy Canada Day.

    ReplyDelete
  11. Prusiner is not a good analogy. There was not a strong competing model for CJD, scrapie, etc. It was simply assumed it was viral. This is not the case with orangutan-human. There is good evidence of an alternative model.

    ReplyDelete
  12. This comment has been removed by the author.

    ReplyDelete
  13. Thanks, John.

    "They cite the paper only to note that it omits characters that would unite Homo with Pongo."

    So, he is basically accusing them of dishonesty... Nice.

    I have the paper somewhere, I'll have to dig it out. I am certain that theyu used over 100 characters. Has anyone seen the Appendix S1 with their list of characters?


    Additionally, I find Schwartz's tone to be nearly on par with what one finds in creationist literature - condescending, dismissive, etc. The section titled "Molecular inconsistency" is a case in point.

    ReplyDelete
  14. The supplemental material is freely available online.

    ReplyDelete
  15. "Has anyone seen the Appendix S1 with their list of characters?"

    That dataset includes some great characters. Among my favourite:
    3. Flexed rectum
    64. Beard and mustache present

    Ladies and gentlemen, we have truly entered the twilight zone.

    ReplyDelete
  16. I was originally neutral about human origins, but became interested when I heard Alan Walker refer to the orangutan theory as the lunatic fringe.

    I made inquiries and found that opponents of the theory could not give me a satisfactory reason other than molecules being right.

    I looked into Schwartz’s morphological evidence and found it to be better supported and corroborated than the chimpanzee morphological evidence.

    I have become intrigued by the possibility that the morphological evidence in support of the orangutan relationship is more correct than the current measure of molecular similarity. It is congruent with the fossil record showing hominids to also share these features (at least the hard tissue features preserved in the fossil record) and early hominids also exhibit other features only found in orangutans and their fossil relatives.

    It is possible that the molecular theory is correct, and vice versa. At present molecular theory will need to give a better response than denouncing the morphological evidence or asserting that molecular evidence is infallible.

    Yes, various morphological studies did leave out, without justification, various human-orangutan characters. I see this as an oversight.

    No, Denis McCarthy is not at the same institution as John Grehan.

    Flexed rectum – this was a feature proposed for humans and African apes by Groves. Not my idea, but included as possibly valid even though not well documented.

    Yes – humans and orangutans are the only primates to have a well developed beard and mustache. A few have a well developed mustache, and some a well developed beard. It’s quite interesting that you can look at an orangutan and see three features found only in humans, and one more usually only in males.

    ReplyDelete
  17. But if you're interested in soft tissue characters just look at the Gibbs et al. (2002, Soft tissue anatomy of the extant hominoids: a review and phylogenetic analysis. Journal of Anatomy 200, 3–49) study that provided support for a Pan-Homo clade.

    ReplyDelete
  18. John Grehan says,

    It is possible that the molecular theory is correct, and vice versa. At present molecular theory will need to give a better response than denouncing the morphological evidence or asserting that molecular evidence is infallible.

    Let me give it a try.

    Evolution is defined as a change in the frequency of alleles over time.

    When you compare the sequences of human, chimpanzee, gorilla, and orangutan genes you inevitably find that the differences between the human and orangutan genes are greater than the differences between human and chimpanzee sequences or human and gorilla sequences.

    Not only that, the sequence differences cluster in a way that allows you to draw a tree of the relationships. That tree is almost always unambiguous. It shows humans, chimps, and gorillas forming a clade that's separate from orangutans. This is true of dozens and dozens of genes.

    The advantage of molecular evolution is that it is much more objective than comparing morphologies. Anyone can look at the sequence data and draw the same conclusions as long as they can distinguish between identical residues and those that are different.

    John, in order to make your case it isn't sufficient to just list your subjective view of the morphological comparisons. You also have to explain why the sequence data conforms to a phylogeny that is very different from the one you propose.

    How do you explain the fact that human and chimp sequences are so much more similar than human and orangutan sequences? Keep in mind that the molecular data by and large deals with neutral mutations so you can't use convergent evolution as an excuse. Also keep in mind that evolution is a genetic phenomenon—it's the genes that count.

    You can't just dismiss the molecular data because you don't like it. You have to offer a realistic explanation for why it could be so wrong. Good luck.

    ReplyDelete
  19. ‘Monkeyman’ refers to Gibbs et al. (2002) support for a Pan-Homo clade.

    This paper is highly problematic as it only uses 'gibbons' (one, all, some?) as an outgroup and the features are taken as read from the literature (often very restricted and many from the 19th century) without documentation of their validity. I notice that ‘Monkeyman’ will use morphometrics to compare with molecular phylogenies. This is also problematic in terms of meeting cladistic requirements. It will be interesting to see the outgroup analysis.

    Larry Moran tries to give a better response than denouncing the morphological evidence or asserting infallibility of molecular evidence, but that is still the essence of his argument. He gives a nice and reasoned summary of many assumptions about the superiority of molecular analyses. I liked the way he stuck to the issues.

    I will give a response to his points to at least show that there is a counter point, that there is an argument to be made about questioning molecular similarity just as there is an argument to be made about questioning morphological similarity.

    “Evolution is defined as a change in the frequency of alleles over time.”

    Yes this is one way of defining evolution.

    True – bean counting of molecular differences show less between human and chimpanzees, but this contains no information about whether such differences are primitive or derived.

    The clustering of sequence differences does allow you to draw a tree, but that does not meant the tree is accurate, whether or not it is ‘almost always’ unambiguous since there is no prior limiting of the differences to those that are derived so it does not matter if the same answer comes from dozens and dozens of genes. A phenetic result is still a phenetic result.

    Although molecular evolution is said to be “much more objective” there is more to this than meets the eye. There is no ‘objective’ way to know what any one base represents as a replacement of a preceding base so one may always identify the base correctly, but its phylogenetic meaning becomes a separate inference.

    Morphological features may be individually ‘subjective’ but they are directly accessible to independent critique and falsification. Perhaps what is most interesting is that despite this supposed problem, the orangutan evidence is so unambiguous.

    As to having to explain why the sequence data conforms to a phylogeny that is very different from the one proposed, one may also argue that it is not sufficient to just list DNA base similarity without explaining why the morphological data conforms to a phylogeny that is very different from what the molecular theorists propose. But we do offer some suggestions in our paper.

    I could argue that it is not for me to ‘explain’ why human and chimp sequences are so much more ‘similar’. But the primary reason for incongruence in phylogeny is the influence of primitive retentions. In morphology there is greater similarity between humans and chimpanzees than humans and orangutans. No reason why the same would not apply to DNA.

    The injunction not to use convergent evolution as an excuse is quite funny given that this is the very ‘excuse’ that molecular supporters use to ‘explain’ away the morphological similarity of humans and orangutans. I am not aware of any empirical necessity for neutral random mutations to track phylogeny.

    Some would argue that evolutionis also epigenetic. But sequence analysis is not ‘genetic’ any more than morphology, it’s about DNA morphology. But the whole point about the morphological analysis is that shared derived morphologies have to also be genetic.

    “You can't just dismiss the molecular data because you don't like it. You have to offer a realistic explanation for why it could be so wrong. Good luck.”

    With a statement like that I presume you have not read the paper. We do offer some ‘realistic explanations’ and sure there will be molecular rebuttals just as there are morphological rebuttals to the molecular explanations. And thanks for the reasoned argument.

    ReplyDelete
  20. John Grehan: True – bean counting of molecular differences show less between human and chimpanzees, but this contains no information about whether such differences are primitive or derived.

    This is a true statement only in the trivial sense that one cannot know anything with certainty when one limits the data to a narrow comparison of the sequence in just two species. It is clearly false in the case where sequence comparisons across multiple species of various degrees of relatedness show an extremely high probability that a particular SNP was a recent event in a common ancestor of a subset of species. In such cases there is LOTS of quantitative information. And there are gobs of genes were we have the sequence data to make robust comparisons. The devil is in the details, and the case for phylogeny based on the molecular data gets stronger as you look deeper at those details.

    The rest of your posting indicates an unwillingness or an inability to confront the molecular evidence head on; essentially ducking the responsibly Larry correctly notes you have: you have to explain the molecular evidence as it exists (in a manner other than just verbal hand waving, which is essentially what you have done here). Not all of us are willing to pay to read the article, so why not post your best arguments here if you have them? What you have said so far doesn’t address the concerns that have been raised.

    Here’s an example of what, IMO, you have to address: in the GULO gene that allows most animals to synthesize vitamin C, most primates instead have a pseudogene due to a frame shift mutation that produced a premature stop codon. In a comparison of the human, chimp and orang sequences in just the 50 base pairs that bracket this mutation, there are 6 locations where these 3 species differ (i.e., at more than 10% of the locations). In 5 of those, the orang is the outlier, and in one the chimp is the odd one. In that one single instance where human and orang are the same, comparison to the sequence in other distantly related species (e.g., cow, macaque, and guinea pig) shows that the chimp is uniquely different.

    The most parsimonious explanation that accounts for all the GULO differences in this region in these three species is that human/chimp line first diverged from the orang line, then the human and chimps split. 3 of the six differences then occurred by simple point mutation in the common human/chimp line, 2 in the orang line, and one in the chimp line. No low-probability reversions or duplicate mutations are required.

    A phylogeny that posits the chimp lineage splitting off first would require either 1) three duplicate mutations and/or reversions in independent lines (to explain 6 differences!), or 2) preexisting multiple alleles that had the different SNP sets, which then coincidentally got fixed one way in the human and chimp lines and the other way in the orang line. #1 is exceeding unlikely. And while #2 is possible, one would also have to propose that there was a similar preexisting variability and the exact same coincidental post-split fixation in the different lineages at all other areas of the genome. That strikes me as also exceedingly unlikely.

    The molecular data is not perfect: low probability duplicate mutations and reversions can occur, preexisting variable alleles can have existed, fixation is a probabilistic process, etc. So one example in one limited region is not dispositive, but can be quite compelling when based on many regions. Certainly it is more quantitative than comparisons of morphological features. It is not sufficient to merely mention vaguely the limitations of a technique (as you are doing here) without showing exactly how those limitations lead to (what you claim are) erroneous conclusion. [We get enough of that from the creationists.]

    ReplyDelete
  21. I should probably clarify that I am not an anonymous blogger (I know Larry gets a little upset about those sorts): my name is Alex Bjarnason, I'm a PhD student at University College London studying morphological evolution of New World monkeys. But I'm also working on great ape and human morphological phylogenetics, particularly in light of the Collard & Wood (2000) and Lockwood et al. (2004) PNAS papers.

    I should also apologise for being somewhat flippant in my earlier posts.

    There is one point that I think is at the very basis of the opposition to molecular phylogenetics exhibited- that sequence alignment is a phenetic method. This is a really very interesting point and one I've thought about before. I think there are two things worthy of consideration-

    a) if alignment is phenetic, we ought to remember the utilisation of an outgroup that gives polarity, or direction, to evolutionary change makes popular molecular methods phylogenetic and not phenetic. The alignment doesn’t create your tree, so utilising a phylogenetic method to generate your tree avoids the phenetic problem IMO.

    b) the alignment procedure is really about connecting homologous structures- is it any more phenetic that selection of Cartesian coordinates in GMM, or indeed choosing morphological characters.... all these process involve connecting structures by their similarity at some point.

    Regarding your comments on Gibbs et al. (2002), I may have misread both papers, but don't you make the same ‘mistake’ by use of a single Hylobates entity (possibly mixed group of multiple species or genera?). I also notice you pool monkeys of the old and new world into a single group, which seems to be a massive assumption that they either form a monophyletic group or don't have enough to diversity to be polymorphic for any of your characters. I do wonder about the wisdom of lumping atelids, pitheciids, cebids with papionins, cercopiths and colobines into one group- that seems terribly simplistic in light of the morphological variation I have seen myself.

    I also find the coverage of the Strait & Grine (2004) lacking, particularly justification for your own dataset over theirs. Their outgroup selection (three outgroups as I recall) also avoided the single outgroup problem you claim Gibbs et al. (2002) had for their soft tissue work.

    The use of geometric morphometrics and phylogenetics is definitely an interesting topic worthy of discussion, but I fear it would prove confusing mixed in with the current discussion. For what it's worth, there are problems, one or two assumptions, but I think also hope of solutions. It is a conversation I would be happy to have with any interested parties.

    ReplyDelete
  22. Response to Alex (all good points that are raised)

    “utilisation of an outgroup that gives polarity…makes popular molecular methods phylogenetic and not phenetic.”

    This is an often stated argument, but it is problematic for molecular analysis where they do not restrict data to those that are individually identified as unique to the ingroup. Where they do the taxonomic representation of outgroup members is so limited as to be potentially very misleading (ie. An outgroup of a few species may not reveal an ingroup base, whereas it might prove to be present if more species were compared).

    “The alignment doesn’t create your tree, so utilising a phylogenetic method to generate your tree avoids the phenetic problem IMO.”

    Not really. Alignment creates homologies as a best overall fit of theorized gaps and substitutions. Analyzing this data using any clustering method (parsimony or other) does not transform the original phenetic homologies into cladistic homologies.

    “ alignment …any more phenetic that selection of Cartesian coordinates in GMM, or indeed choosing morphological characters....”

    In a general sense yes (sorry – I’m too old to learn shorthand such as GMM, whole English words work best for me), but as above the homologies in alignment are a compromise best fit of all potential homologies. In morphology the homologies can be hypothesized individually.

    “…Gibbs et al. (2002)…don't you make the same ‘mistake’ by use of a single Hylobates entity (possibly mixed group of multiple species or genera?).”

    The difference is that we do not restrict our outgroup to just Hylobates. In principle all Hylobates species and at least all OW monkey species comprise our outgroup (not just a few selected species as in many molecular and morphological studies).

    “… you pool monkeys of the old and new world…assumption that they either form a monophyletic group or don't have enough to diversity to be polymorphic….”

    No we do not assume their monophyly. In pooling we are arguing that all species of NW and OW monkeys are the actual or potential outgroup (we note in detail for some characters that the outgroup condition is not known for some or many species)

    “… wonder about the wisdom of lumping atelids…colobines into one group...”

    As above, our point is that ALL species should be considered for the taxa comprising the outgroup. In some cases the documentation is very good in this respect. For example, the incisive fenestra is probably known to be present in all 250+ species of monkeys and hylobatids, and the incisive foramen of the ingroup is a therefore an extremely robust apomorphy.

    “I also find the coverage of the Strait & Grine (2004) lacking, particularly justification for your own dataset over theirs.”

    Not sure I understand this sentence. The justification of our dataset is that it encompasses many more outgroup species.

    “Their outgroup selection (three outgroups as I recall) also avoided the single outgroup problem you claim Gibbs et al. (2002) had for their soft tissue work.”

    Its not the ‘single outgroup’ that is the problem, but the range of the outgroup. Our view is that restricting the outgroup to a few species of hylobatids is inadequate. The outgroup for Strait and Grine is also inadequate as it is restricted to Hylobates and two monkey genera (and even these comparisons show that some of their ingroup features are found in one or more outgroup taxa).

    “The use of geometric morphometrics and phylogenetics is definitely an interesting topic worthy of discussion, but I fear it would prove confusing mixed in with the current discussion.”

    I view geometric morphometrics as problematic for phylogenetic reconstruction as they tend to represent overall similarity that may be misleading. I’ve seen linear parameters given for teeth of humans and chimps, for example, that ignore the fact they are otherwise morphologically incongruous with respect to the orangutan.

    ReplyDelete
  23. "Yes – humans and orangutans are the only primates to have a well developed beard and mustache."

    Beard and mustache? Why didn't you say so before! Clearly DNA is useless...

    ReplyDelete
  24. Larry:
    "It shows humans, chimps, and gorillas forming a clade that's separate from orangutans. This is true of dozens and dozens of genes. "

    It is also true of whole mitochondrial genome analyses.

    ReplyDelete
  25. Divalent said you have to explain the molecular evidence as it exists (in a manner other than just verbal hand waving, which is essentially what you have done here). Not all of us are willing to pay to read the article, so why not post your best arguments here if you have them? What you have said so far doesn’t address the concerns that have been raised.

    I think the best way to address this is to read the original paper and the section presenting our points on molecular analysis. I can send a copy if contacted at jgrehan@sciencebuff.org

    ReplyDelete
  26. Divalent argued that the most parsimonious explanation that accounts for all the GULO differences in this region in these three species is that human/chimp line first diverged from the orang line, then the human and chimps split, and that other alternatives are unlikely.

    There are lots of sequence analyeses. I have not looked at them all, but alway interested to take a close look. If one accepts the premise that base similarity is accurate for phylogenetic reconstruction then it follows that the most parsimonous arrangement is a predictor of phylogeny. The problem is the morphological incogruence that suggests the molecular assumption is problematic.

    I would be interested in the citation of the paper referred to so I can take a look.

    I did find Ohta & Nichikimi (1999) but they do not use the sequence to produce a phylogeny, but use the 'accepted' phylogeny to interepret their data (which had to be aligned along with computational corrections on substitution numbers.

    ReplyDelete
  27. Divalent noted that the most parsimonious explanation that accounts for all the GULO differences is that human/chimp line first diverged from the orang line, then the human and chimps split. exceedingly unlikely. Alternatives are considered extremely unlikely.

    There are lots of DNA studies and I have not read but some. But always interested to look further. I would be interested in the citation of the paper referred to by Divalent.

    I did come across Ohata & Nishikimi 1999 but they did not use the sequence to produce a phylogeny, but used the ‘accepted’ phylogeny to interpret their GULO data.

    ReplyDelete