Ancestral sequence reconstruction at each node of a gene tree

Question

[biomonte]

Hi,

Given a (large) gene tree and its corresponding alignment, is there any way to reconstruct ancestral sequence at each node?

In addition, would it be possible to know exactly for each gene in the tree which is its ancestral node (the one that produced the most recent split)?

The tree was reconstructed using homologous genes descending from a single gene in the LCA of all the species being considered, thus it includes both orthologs and paralogs. However I think this should not impact the reconstruction of ancestral sequences.

Thanks

[Jesse]

I vaguely know of one way, at least, if you're working with a maximum likelihood tree: Felsenstein's Pruning Algorithm. That requires that you have a model for the probability of seeing a particular outcome at one node in the tree based on each possible case for the parent node. In that example you have three possible characters at one position and you work backwards from the tips (with 100% "probability" for the tips since you observed them directly) to the root (fuzziest probabilities). I've only dealt with this by using someone else's Perl implementation for an immunoglobulin-specific case, though, and I don't have a great grasp of the math. That page also mentions geiger and PAUP* but I've never dealt with those.