Likelihood versus Parsimony at Long Branch Lengths

In this exercise you'll make a small tree that can be difficult to reconstruct with phylogeny because it has two long brances separated by a short branch. Like this:

    B      D
    |     /
    |    /
    |   /
    |  /
    | /
    -
   / |
  A  C

You're going to evolve sequences on this tree with seq-gen and see which method: parsimony or likelihood better deals with this kind of tree.

1. First make the tree. You can do this really simply with the following UNIX:
   
   % echo "(((A:0.2,B:1.2):0.2,C:0.2):0.5,D:0.7);" > quartet
   
   Can you see why this tree matches the shape drawn above?
   
   
2. Evolve some proteins on the tree. Find your seq-gen executable. Put it in the same directory as quartet. Then execute:
   
   % ./seq-gen -mJTT quartet > q.plp
   
   
3. Compute a parsimony tree. Use eprotpars on q.plp to compute a parsimony tree.
   
   % eprotpars
   
   
4. Install Phyml Phyml is a very fast likelihood program, find the executable version on the net and put in the directory with your data. Run it with
   
   % phyml_macOSX
   
   and make a likelihood tree for this data.
5. Visualize the two trees with TreeView. Which one (or both) is correct?