Darwinian Phylogenists Do the Funky Chicken
Fredrik Ronquist is active in phylogenetic systematics, the art of drawing evolutionary trees from DNA comparisons. And he admires Joseph Felsenstein, an “icon in the field.” But when he reviewed Felsenstein’s new book, Inferring Phylogenies (Sinauer, 2004) in the Feb. 5 issue of Science,1 he had mixed feelings about the author’s biases and his choice of humor.
Ronquist has much to praise about the iconic master’s work, concluding “I can think of no one who could provide a better and more comprehensive summary of the current methods for building evolutionary trees.” Nevertheless, his criticisms are revealing about the state of this art:
- What is it about, anyway? The book seems to omit a rather important part of phylogenetic systematics:
What I found most surprising about the book is that it is not at all about systematics. Readers will find no coverage of many basic concepts in phylogenetic systematics–such as synapomorphy, symplesiomorphy, sistergroups, outgroups, and monophyly.
While Felsenstein covers many subjects like “techniques for statistical testing of evolutionary trees,” uses of phylogenies, and “nearly every quantitative approach to tree-building that has been tried,” Ronquist is most surprised there is no coverage of these important terms and concepts in a 684-page definitive treatise by an expert in the field.
- No help on classification.
Another topic that many phylogenetic systematists consider important but the book glosses over is how one should convert phylogenetic trees into classifications of organisms. According to Felsenstein, “The delimitation of higher taxa is no longer a major task of systematics, as the availability of estimates of the phylogeny removes the need to use these classifications.” Even a cursory look at the literature would prove that many active systematists disagree; indeed, the discussion of classification and naming principles seems to be as vigorous as ever. This neglect of the classification issue is all the more remarkable because Felsenstein devotes an entire chapter–one of the more original and important contributions in the book–to the drawing of trees (specifically, to algorithms for drawing diagrams of trees). After all, drawing trees is just another way of communicating the results of a phylogenetic analysis. Often a diagram is better, but sometimes a name is necessary. I do not think we will ever see papers with titles like “The biology of <insert tree drawing here>.”
- Controversy is bitter.
In a field that has been plagued by outrageously bitter controversy, the book is remarkably balanced on the whole. For example, consider Felsenstein’s summary of the debates on statistical inconsistency. It reveals when parsimony is inconsistent and suffers from “long-branch attraction,” but it makes no secret of the fact that likelihood methods can also be misled by similar phenomena when the model used for inference is incorrect. The attempt to provide balanced coverage probably will not stop ardent parsimony advocates from being disappointed.
This is because Ronquist feels Fenselstein was unfair in his choice of algorithms to exalt, and ones to ignore.
- The Bayesian Funky Chicken. “The book’s coverage of Bayesian inference of phylogenies is surprisingly short and critical,” Ronquist complains. Bayesian inference is a fancy mathematical form of educated guessing by applying values to likely causes, but it suffers from GIGO: garbage in, garbage out. Ronquist disagrees with the author’s criticisms, and is not amused by the joking description Fenselstein gives to this technique as applied to evolutionary tree-building:
I find the author’s complaints about prior distributions partly misdirected. For instance, Mau and Newton’s prior on clocklike trees is described as “technically inadmissible” and “an impossible distribution” because it is an improper probability distribution. But improper priors are often unproblematic in Bayesian inference, and there is an entire school of “objective Bayesians” who routinely use them. The comment on the choice of proposal distributions is more funny than helpful: “At the moment the choice of a good proposal distribution involves the burning of incense, casting of chicken bones, use of magical incantations, and invoking the opinions of more prestigious colleagues.”
After all this reluctant criticism, Ronquist manages to find something to compliment, in closing:
Although it is easy to criticize a book that tries to cover so much, in this case doing so is like throwing stones in a glass house. Every phylogeneticist can probably find some points they understand better than Felsenstein, but I can think of no one who could provide a better and more comprehensive summary of the current methods for building evolutionary trees. It will be a long time before there will be a comparable book; perhaps the field is now growing too fast for there to ever be one. The publication of Inferring Phylogenies is a milestone for evolutionary biology in general and phylogenetics in particular.
1Fredrick Ronquist, “Phylogenetics: A Broad Look at Tree-Building,” Science Volume 303, Number 5659, Issue of 6 Feb 2004, pp. 767-768.
Sometimes you have to just stand back and let the Darwin Party members do it to each other. Does anyone have confidence in evolutionary tree-building after this indecent exposure? When an expert in the field omits significant parts of the story (why? because he feels they are invalid?), characterizes it as a battle over the most prestigious authorities, and describes one of the chief methodologies to be as mystical as casting chicken bones and using magical incantations, what are we supposed to conclude? Don’t they realize it’s confusing to the peasants when the shamans are exorcising one another?
For more on phylogenetic tree-building, see 11/26/2002 and 06/13/2003 entries, and follow the Chain Links on “Genes and DNA.”