Sunday, August 05, 2012

Understanding Phylogenetic Trees

A few months ago I posted an exam question from my course on evolution [Exam Question #1]. It was designed to test student's understanding of phylogenetic trees—a serious problem in evolutionary biology. That post generated quite a few comments.

Take a look at this figure from a 2008 issue of Nature. Do you see the problems?

If not, read David Morrison's guest post on Scientipoia: Ambiguity on Phylogenies. He patiently explains all the common misconceptions about phylogenetic trees and references all the important articles in the scientific and pedagogical literature. Once you read his article you will never look at trees the same way again.

You will also be astonished at how bad the scientific literature has become when it comes to explaining phylogenies based on these trees. You would think that evolutionary biologists would have long ago stopped thinking about directions in evolution but it's surprising how often the great chain of being creeps into modern scientific papers.

[Hat Tip: Mike the Mad Biologist]


  1. Did you happen to see the new textbook (on sale at the Ottawa meeting), Tree Thinking, by David Baum and Stacey Smith? I bought a copy at the incredible meeting discount, and a course based on that book might just help a bit.

    That tree does have one salutary role: it demonstrates that the scala naturae (or, to give it the real cause, anthropocentrism) dominates choices for genome projects, so far.

  2. I have often wondered how you can falsify a claim made about a particular phylogenetic relationship. There doesn't seem to be one given the existence of convergent or parallel evolution, not to mention many variables such as generation span, mutation rates, biased recombination, horizontal gene transfer and so on.

    1. Falsification is the wrong framework for any scientific inference that involves probabilities of events, where no event is absolutely impossible on any hypothesis but their relative probabilities differ between hypotheses. While hypothetico-deductive frameworks for phylogenetics were popular in the 1980s, they have lost ground steadily since then to statistical frameworks where it does not make sense to talk of falsification.

      Similarly, if you are tossing coins, you can't ever falsify the hypothesis that the coin has probability of heads 0.3, even if you toss 100 times and get 60 heads.

    2. Joe, thank-you for saying that.

      There still seem to be a great many scientists out there who think that falsification is the main characteristic of good science.

    3. @Joe: I disagree strongly - evaluating model fit and moving from less realistic to more realistic models are inherently falsificationist in nature. See this paper by Andrew Gelman
      and several discussions on his blog for more...

  3. "Similarly, if you are tossing coins, you can't ever falsify the hypothesis that the coin has probability of heads 0.3, even if you toss 100 times and get 60 heads."

    This depends on what you mean by "falsify". If you are willing to propose a null hypothesis and use a p-value cutoff, of course, you can falsify such a hypothesis in a reasonable sense of the word.

    The post was good but it's not about "ambiguity in phylogenies", which would be about representing uncertainty in topology or branching. It's about "misinterpretation of phylogenies".

    I got to review a draft of Baum & Smith's book, it's quite good.

    1. It's about "misinterpretation of phylogenies" due to the "ambiguity" of the manner in which many phylogenies are constructed.

      Though yes, the title itself is a bit ambiguous as to what the post is going to actually be about, which is rather amusing.

  4. I looked at the article and was immediately thrown off by the author's mis-named distinction between "variational" and "transformational" evolution. Yes, obviously, people bring a lot of baggage to the way that they view trees, and the way that they draw trees, but the implicit message in this blog and in the blog that explains "transformational" evolution is that there are NO legitimate scientific questions that pertain to these issues, only illusions and quirks of thought. Apparently the author assumes that evolution is a process that cannot conceivably produce systematically unbalanced phylogenies, or exhibit directions, or exhibit goal-directed behavior. To me this sounds like another case of evolutionists throwing out the baby with the bathwater in a desperate attempt to establish orthodoxies. Is "Coynian" an adjective?

    I would have had more sympathy if the author had just stated clearly "under assumptions X, Y and Z, which I explicitly posit as a null model, phylogenies will be balanced and will show no signs of trunks or crowns or ladders, etc, and we should agree to imply trunks, crowns, etc ONLY when there is evidence against the null model".