Ahmed Moustafa wrote:
> My question is why not rather than performing the multiple sequence
> alignment, simply use pairwise alignments scores to fill in the
> distance matrix and from there to construct and infer the tree.
I think that clustalw works like this, doesn't it?
This is because there are no real good multiple alignment sequences that can
tackle more than a few dozen sequences at once without an explosion of time
and memory requirements. (Stoye et al have some nice publications on this).