summary of guidelines for phylograms

Yannick at CX13 CX13 at MUSICA.MCGILL.CA
Wed Feb 12 11:01:14 EST 1992

OK people, thank you for responding with such useful information
to my original request (see below) for guidelines on generating
phylograms. As you can see, the answers are pretty consistent. The
response shows just how useful something like to mol-evo BBS really
is. Thanks!


In article <03FEB92.24040546.0081.MUSIC at MUSICA.MCGILL.CA>
 ("Yannick at CX13") writes:

>From: CX13 at MUSICA.MCGILL.CA ("Yannick at CX13")
>Newsgroups: bionet.molbio.evolution
>Subject: Guidelines on tree-making?
>Message-ID: <03FEB92.24040546.0081.MUSIC at MUSICA.MCGILL.CA>
>Date: 4 Feb 92 03:15:35 GMT
>Sender: daemon at genbank.bio.net
>Distribution: bionet
>Lines: 21
>G'day fellow netters.

>I have what is probably an ignorant question regarding
>phylogenetic trees. I have noticed that people publishing trees

>often publish trees by two methods: one tree produced by a
>distance method and another produced by, say, parsimony. I
>presume this is a way of confirming their results by showing that

>different methods generate similar or identical trees. Is that
>the case? ALso, is that really necessary, or can you rely only on

>a parsimonious tree?

>Lastly, can somebody point me toward some published guidelines
>concerning generating phylogenetic trees? I have been to short
>courses and have read textbooks, but I feel pretty ignorant


>Yannick Pouliot
>Montreal Neurological Institute


Hi Yannick;
     I saw your posting on BIONET.GENERAL asking for reviews on
phylogenetic analyses. There are many methods for phylogenetic
analysis (e.g. Parsimony. Distance Matrix Fitting, Character
Compatability, Maximum Likelihood, Invariants); they have been
developed separately over thew last thirty years, and congruence
of results between different methods is seen as a good but not a
necessary thing. Each method makes its own assumptions about the
way evolution works; your trust in a method of analysis ultimately
comes from how much you believe in the model of evolution which
that method embodies. Even though I am a fan of of Parsimony
methods, I would reccomend that you read wider in the literature
to get a feel for what is going on. Try the follwoing references:

   Felsenstein, J.~S.
   \newblock Phylogenies from molecular sequences: Inference and

   \newblock {\em Annual Review of Genetics}, 22, 521--565, 1988.

   Swofford, D.~L. and Olsen, G.~J.
   \newblock Phylogeny Reconstruction.
   \newblock In D.~M. Hillis and C. Moritz (eds.) {\em Molecular

       Systematics}, Sinauer Associates, Inc., Publishers,
       Sunderland, MA, 1990. 411--501.

Swofford and Olsen (1990) is the VERY, VERY BEST review I have yet
- Todd

Todd Wareham   harold at odie.cs.mun.ca
Department of Computer Science
Memorial University of Newfoundland
St. John, NF, Canada   A1C 5S7



I suggest that you find the new book, "Phylogenetic Analysis of
DNA Sequences" edited by M.M. Miyamoto and J. Cracraft. Also
the chapter by Swofford D.L. & G.J. Olsen. 1990. Phylogeny
reconstruction. Pp. 411-501 in "Molecular Systematics (D.M.
Hillis and C. Moritz, eds.). Sinauer. is helpful. The first
book was from Oxford, and you might ask if they will still
offer you their 20 % discount (says till 1/31/92: $39.96 w/
discount). I sent in the part with the address and phone #.

Personally, I see no reason to present a distance-based tree,
as there is typically reliance on the assumption of constant
rates in the different branches, as well as reliance on
similarities that are irrelevant to supporting the monophyly
of particular clades (i.e., plesiomorphic characters). If
this sounds foreign, perhaps start with E.O. Wiley. 1981.

Doug Eernisse
usergdef at ub.cc.umich.edu                 |
usergdef at umichub.bitnet                  |
| Museum of Zoology, Univ. of Michigan, Ann Arbor, MI 48l09 USA

Yannick Pouliot,

     I have just submitted a phylogeny paper where we took a pile
of gene sequences and made trees from them.  My particular bias
about tree making is to try to "approach" or gain more confidence
into the best tree you come up with (try to estimate the "correct
tree").  The best way to do this is to use  many of the different
methods (max. likelihood, neighbor joining, wtd. parsimony, max.
parsimony, and others).  Read the wonderful review from Swofford
and Olsen in the book Molecular Systematics; it goes into the
various assumptions each method makes.  At present, it is no more
correct to choose a parsimony tree over a neighbor joining tree if
you do not know many specifics about the sequences you are studying
(like codon bias, or unequal rates of mutation along gene lineages,
etc, etc)  In many cases certain methods have been shown to
converge on  an incorrect tree because of violations of the
assumptions.  Hope this helps!

mike strong

mstrong at uci.edu

More information about the Mol-evol mailing list

Send comments to us at biosci-help [At] net.bio.net