James McInerney wrote:
> Hi,
>> I am having a problem trying to unambiguously reconstruct ancestral
> sequences. I know it is possible to resolve ancestral sequences _with_
> ambiguity, but what about making some educated guesses and/or
> assumptions and then deciding to construct unambiguous ancestral
> sequences (sort-of like maximum likelihood). We can choose among
> competing alternatives using a pre-constructed substitution matrix to
> choose which ancestor might be the most likely. The problem we are
> finding arises when we get to the root.
>> Is there any method that has been published that will succesfully
> reconstruct ancestral sequences based on a tree, without having to
> resort to introducing ambiguous residues?
This cannot be done. For complex data sets one can assign multiple
ancestral sequences to the earliest node and obtain equally parsimonious
trees. There is simply no way to decide among them. That is why Fitches
method has so many ambiguous assignments.
Mike Syvanen
---
