phylogeny programs

Mary K.Kuhner mkkuhner at kingman.genetics.washington.edu
Mon Aug 13 18:10:53 EST 2001


In article <9l9hj2$qid$1 at mercury.hgmp.mrc.ac.uk>,  <rodgers at onramp.net> wrote:
>I'm interested in "succint" comparison/critique of the various
>algorithms/programs available for phylogenetic-type sequence analyses.
>I understand this may be like asking for a "succint" critique of
>Mozart, but still... 

A useful web site is the course site for the Molecular Evolution
workshop at Woods Hole:

http://courses.mbl.edu/Courses/mole.html

This contains lecture notes from a two-week intensive workshop
on topics like this one, many of which have very good reference
lists.  For your particular questions, try the lecture by Joseph
Felsenstein on bootstraps and other means of assessing tree
reliabilities, and the lecture by Paul Lewis on comparing different
phylogeny methods.

>    In particular - is there a way to assess the "error" or confidence
>level in a tree?  I understand there are some outputs that indicate
>the percentage of bootstrap trees that give the final tree shape.  

The program package PHYLIP  and the program PAUP* can both
carry out a bootstrap analysis, and have good documentation on
how the method works.  Have a look at their web pages.  PHYLIP
is free; PAUP* has more features and a slicker interface; both are
pretty reliable.

Different tree estimation methods also come with their own
internal tree-support measures.  For example, likelihood methods
can offer approximate confidence intervals on the length of a branch.
If the confidence intervals overlap zero, we know that we are not
even confident that the branch exists at all.  This is also implemented
and discussed in PHYLIP and PAUP*.

>   Also- how does one interpret the vertical and horizontal distances
>in rooted dendograms as, for example, output from Treeview?

The distance along a branch from root to tip is often drawn
proportional to some measure of (possibly corrected) evolutionary
distance.  Sometimes trees are simply drawn as topologies with
arbitrary branch lengths, so watch out.  If the tree is in a published 
paper, they should indicate the scale.

The distance on the page from one tip to another is almost
always meaningless.  You can rotate parts of the tree to bring
tips closer together or further apart without having any effect
on the tree at all--it's just a presentational issue.  (The less
scrupulous among us may prefer to order the tips in such a way
as to give more visual support to their preferred groupings!)

Mary Kuhner mkkuhner at genetics.washington.edu

---





More information about the Mol-evol mailing list