Skip to main content

Archived Comments for: Rooted triple consensus and anomalous gene trees

Back to article

  1. Species tree estimation

    Scott Edwards, Harvard University

    17 June 2008

    I found this article very interesting. A clarification, however: the Bayesian model developed by Liu and Pearl (Syst. Biol. 56, 504 ,2007) and modified by Edwards, Liu and Pearl (PNAS 104, 5936, 2007) can accommodate variable population sizes within lineages. Thus the statement in your article that one drawback to the Bayesian (BEST) method is that it cannot deal with situations in which "...the population parameter will likely not be constant in time or across species" is not correct. In the version 1.6 of BEST, available on the web, one can estimate the posterior distribution of the population parameter theta for each node in the species tree separately, although the prior for each node is the same. In fact, analyses producing different posterior distributions for theta have been published in the original descriptions of the model. A multiple-allele model that can also measure variable population parameters among current (in addition to ancestral) species is in press in Evolution.

    The point about the computational burden of this (and likelihood) methods being prohibitive is indeed correct. Liang Liu has developed a BEST version 2.0 (also available on the BEST web site) that is considerably faster than the previously posted version and involves only a single MCMC algorithm. Additionally, your point about variable rates of evolution among lineages (molecular clock hypothesis) is also correct. We are currently developing an extension of the Bayesian approach, as well as other parametric approaches, that will circumvent this problem.

    Thank you for your attention to the problem of multilocus species trees, which I think is one of the most pressing in phylogenetics and phylogenomics today. There will be a symposium at the Evolution meetings in Minneapolis in late June on this very topic.

    Competing interests

    None declared

Advertisement