drawing trees

[Frank Wright]

Hi All,

Gene Cutler asked:

>Hello, all.  I have a question about phylogenetic-type trees for 
>sequences.  I haven't quite figured out how to do this using 
>emboss/phylip.  This is how I have been doing this with gcg:
>
>run gcg program distances on the msf file
>run gcg program growtree on the distances file
>
>How would I do this with PHYLIP instead?

The GCG DISTANCES program and GCG GROWTREE programs are very similar to
the DNADIST/PROTDIST and Neighbor programs in PHYLIP.  In other words,
they allow phylogenetic trees to be constructed using "distance-based"
methods, but do not allow maximum likelihood or parsimony methods to be
used.  They also don't do bootstrapping tests, tree comparisons, and
lots of other things.

If you are using distance-based phylogenetic methods, some notes:

(1) Weighted least-squares is slower but more accurate than
Neighbor-Joining, so use the PHYLIP FITCH program instead of NEIGHBOR.

(2) Recently, the PHYLIP-like WEIGHBOR program (Weighted
Neighbor-Joining) has been released.  WEIGHBOR appears to be an
improvement on Neighbor-Joining (and possibly weighted least squares). 
See http://www.t10.lanl.gov/billb/weighbor/.  I've not tried it out much
but the simulations in the published paper look convincing.

(3) PHYLIP (version 3.6) has improved DNADIST (more distance methods)
and PROTDIST (rate heterogeneity among sites added).

Best Wishes,
Frank
-- 
Frank Wright
Biomathematics and Statistics Scotland, 
SCRI, DUNDEE DD2 5DA, Scotland
frank at bioss.sari.ac.uk