In article <9212101644.AA15862 at t13.lanl.gov> asl at T13.LANL.GOV (Alan Lapedes) writes:
>I am interested in obtaining one or more phylogenetic tree codes that implements
>a conventionally accepted algorithm for producing trees. Apparently,
>phylogenetic tree algorithms seem to be a somewhat controversial topic -- hence
>my request for code that, even if not universally accepted, has at least been
>analysed and debated in the literature.
>>(1) Would prefer C, Fortran is OK.
>(2) Would like to be able to produce not only trees, but also a
> parsimonious reconstruction of the sequences at the nodes.
>(3) Would like to be able to handle either nucleotides, or amino acids as input.
>(4) Would particularly like to be able to handle a few hundred amino acid
> sequences, up to a couple of hundred residues long as input.
>(5) It's critical to us to have reasonable speed: in our admittedly naive hands
> the "protpars" code for a problem the size of point (4) above takes too
>(6) If the code is in C or Fortran we could possibly run it on supercomputers
> for a speed-up. If the code is already ported to vector or parallel
> machines we would appreciate information on acquiring the code, or accessing
> the machine (if a service is provided).
Our automatic server implements a basic tree (Felsenstein's algorithm)
with several optimizations. People are generally happy with the trees
(1) - it is written in Darwin, which is written in C.
(2) - not for all nodes, only for the root of the tree
(3) - only peptides
(4) - no problem, we have tried it with more than 500 sequences
(5) - Depends on what you consider fast. For a tree with n sequences
most of these algorithms are O(n^3) so it depends heavily on
the number of sequences.
You can make up your own mind on the quality of the trees and
on the speed by using the automatic server. Send a message with
"help AllAll" as body to cbrg at inf.ethz.ch. If you like it, you can
obtain your own copy of Darwin.
Gaston H. Gonnet, Informatik, ETH Zurich.