[Taxacom] DNA contamination

[Karl Magnacca]

On Fri, April 1, 2011 8:13 am, Lynn Raw wrote:
> I was wondering if contamination is a more general problem affecting
> molecular phylogenies. It might explain one that I saw with an
> African species nested within a set of apparently unrelated North
> American species.

Contamination is a problem with things like human sequences getting
into primates, fungi in plants, and some other circumstances, but in
most cases it's relatively detectable; unless you end up with a
complete fungal phylogeny, you're going to have a really long
branch.  I think mislabelling sequences is an even bigger and much
more insidious problem, especially in multigene phylogenies.

One paper I reviewed had two species from the same genus separated
by relatively long branches.  This isn't necessarily a red flag,
since the genus was big and there were only those two.  In response
to some other concerns I had, in the revised version they included
separate trees for the two genes - for one, those two species were
nearly identical, and in the other, they weren't even sister
species.  It turned out the sequence for one gene had been swapped
with another species.

Likewise, I had to deal with an older dataset that consisted of five
concatenated genes, and in each gene set at least two pairs of
sequences were obviously swapped; there may have been more that
weren't so obvious because they didn't involve species from
different groups.

The upshot - even aside from actual DNA contamination, you have to
be careful in making sure your sequence names really come from the
right extractions!

Karl
=====================
Karl Magnacca
Postdoctoral Researcher
University of Hawaii-Hilo