[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Subject Index][Author Index]

Re: Bayesian phylogeny



ekaterina A wrote:

> I would like to know if any one of you all  on this
> list (Mickey Mortimer/Jaime Headden?) have used
> Bayesian inference for reconstruction of phylogenies
> using the same data matrices that you use for the
> parsimony analysis.

Bayseian approaches to phylogenetics for discrete, non-molecular data are
currently unavailable to the public at large. This situation is directly
related to the reason that maximum likelihood is not currently in use for
the same data: both use the likelihood function. Although there is currently
an (apparently) acceptable model for likelihood using discrete data, in the
following paper:

Lewis, Paul. 2001. A Likelihood Approach to Estimating Phylogeny from
Discrete Morphological Character Data. Systematic Biology, 50:913-925.

   ... contrary to the paper, there is no available program that
incorporates this model. I contacted Lewis, and he told me it will be
available in PAUP* 4.b11 and MrBayes 3.0
    It is actually possible to wheedle MrBayes 3.0 from John Huelsenbeck,
but I have not done so yet. Students in Paul Lewis' group, and John
Huelsenbeck and his collaboraters, are examining the potential ramifications
of this model for likelihood and Bayesian methods (respectively), and they
have presented preliminary results at meetings. I actually recently missed a
workshop with Huelsenbeck in which he discussed approaches to placing
stratigraphic information in a Bayesian analysis.

      I myself am actively interested in using these techniques for
paleontological research. I encourage any and all of you to explore their
potential. I agree that these methods are promising, although I caution you
that they may aggravate some old wounds (e.g., "branch lengths" and the
informativeness of apomorphies). I recently discovered that it is NOT
advisable to discuss these techniques with veterans of the original Clade
Wars, or you'll hit some serious post-traumatic stress symptoms!

    And finally, a word on programs: Because of the way likelihood-based
techniques work, it is not advisable to simply convert a morphological
matrix to ACTG nucleotide format to run it. If you want to try these
techniques, get the appropriate software (MrBayes 3.0 or PAUP*4.11b). I
wouldn't be surprised if Felsenstein produces a module for Phylip in the not
too distant future. Also, if you wish to learn more about these techniques,
you will have to learn about how likelihood is used for molecular data
first, then learn about Bayesian methods in the same context. I recommend
starting with the Hillis et al. moelcular systematics book, which is a
resource ANY serious systematist should acquire (no, I can't afford it,
myself).

    Hope this helps!

    Wagner