[bip] Interesting article on multiple sequence alignment

[Erich Schwarz]

On Thu, 24 Jul 2008, Jongmin Nam wrote:

> Perhaps it's good time to think about incorporating indels in
> sequence analysis (e.g. distance estimation, inferring ancestral
> sequences), as their method seems better for finding the true
> history of insertions and deletions.

    In cases where orthology is clear and not too messy, this could
make a lot of sense.


> One of the remaining questions is how practically easy this
> approach would be when we don't know the true phylogenetic
> relationship of the sequences.

    I think this method would be impossible to practically apply to
a complex protein family with multiple paralogies.  It would work
best for protein families where essentially only 1 protein-coding
gene exists per species, and where only 1:1:...:1 ortholog
relationships between the genes of different species exist, so that
figuring out the species tree automatically constrains the gene
tree.

    The basic idea, however, that phylogeny and sequence alignment
should be worked out in parallel is very old: it was first proposed
by Sankoff in the 1960s, if I remember correctly.  Joe Felsenstein
covered that (and much related stuff) at great length and depth in
his recent monograph:

    http://www.sinauer.com/detail.php?id=1775
    http://www.amazon.com/Inferring-Phylogenies-Joseph-Felsenstein/dp/0878931775


--Erich