[Molecular-evolution] Re: Bootstrap interpretation
Mary K. Kuhner
(by mkkuhner from kingman.gs.washington.edu)
Tue Jun 10 14:00:10 EST 2008
In article <mailman.370.1213016647.3533.mol-evol from net.bio.net>,
DimitryASuplatov <genesup from gmail.com> wrote:
>I want to generate "random" alignment to use it as a neutral evolution
>model for hypothesis testing. I need a phylogenetic tree of my species
>I have used phylip package to generate a set of 1000 replicates, then I
>have used PMB model to calculate the distances with protdist and finally
>appllied neighbor and consese with default settings.
>As a result a have an unrooted tree with bootstrap values generaly
>better at the tips than in the center of the tree. In my understanding
>this is a normal case, though some values are equal to 25%.
>So my question is what could I do with this tree? As I understand, 25%
>values emphasize the fact that different parts of distant sequnces do
>not show consistency with overall comparison. I can not crop the
>alignment. Should I use neighbor-joining to calculate the tree directly
>from the alignment?
I am having trouble understanding what you want to do here. Let me
try to restate it.
If I understand correctly, you would like to simulate random data on
the same tree that generated your species. You are trying to get
this by first making a bootstrap consensus tree of your species, and
then (you don't describe this part) simulating data on that tree.
This won't work, because the bootstrap consensus tree has no branch
lengths, and you need branch lengths to simulate data. The usual
approach is simply to infer the tree from your original data (by
whatever method you prefer) and simulate data on that inferred tree.
While the inferred tree will not be perfect, it should be generally
similar to the unknown true tree and may be good enough for a
If this is not what you are trying to do, could you restate it?
Mary Kuhner mkkuhner from u.washington.edu
More information about the Mol-evol