Redrawing the family tree

[redbird]

[rysmiel, livredor, does this seem reasonable and/or startling to you?]

I get the New Scientist rss feed here; this morning, they tossed us another redrawing of the mammalian family tree, which puts horses closest to Carnivora, and close to bats, with cattle closer to whales than horses.
Before cheerfully putting horses closest to carnivora and a distance from cattle, the authors note that

Comprehensive analyses of large collections of DNA sequences mostly reject the conclusions from morphological analyses.

I'm not sure what that does to information derived from fossils.

Reassuringly, marsupials are still an outgroup to the rest of the mammals; the authors seem not to have tested sequences for monotremes, or at least didn't include them on their chart.

Comments

[doseybat.livejournal.com]

I do not know much about animals but my PhD involves DNA based evolutionary tree reconstructions for plants, and comparing the DNA results to morphological data. In botany molecular studies have overthrown many traditional morphology derived views. It turns out a lot of things look similar because of convergent adaptive evolution, and while for the majority of groups the way they look is a good predicitive factor, for some it is really not. So I am not really surprised by the horse/cattle (Oopswordlimit)

[aquaeri.livejournal.com]

I'm in molecular evolution and have been vaguely following the redrawing of the mammalian tree over the last 10-20 years. It's been pretty significant, with Laurasiatheria and Afrotheria both entirely DNA-based.

As far as I understand (not being a fossil person) there aren't enough fossils that are well-enough understood that this sort of thing upsets all that much.

As an earlier example, it used to be thought that humans diverged from other primates about 15 million years ago. The molecular/DNA date is more like 5 million. This meant that all the fossils between 15 and 5 million years which had been thought to be on the uniquely human lineage had to be reexamined. Some of them have been moved over to the orangutan lineage! No-one is unhappy about this in retrospect, although it caused outrage at the time.

I'm afraid if you didn't know that whales and cattle are quite closely related, you're just behind the times :-). I don't know anyone arguing against that grouping any more.

The crunch in the paper is basically whether or not you agree that the four events that support (bat, horse, carnivore) to the exclusion of cows and whales are unique or not. There's one conflicting event, so despite their claim about how fabulous the technique is, conflicting events obviously do happen. And that conflicting event just happens to support the previous tree (by their explanation for how this event arose, it's 50-50 odds.)

I think their data is interesting and suggestive, but I'd like more, so I can get an idea of how frequent conflicting events are, and thus how many events on the current tree might also be. And I don't think much attention should be paid to their reanalysis of Murphy et al., as there's essentially no statistical support to favour any proposed tree over any other, no matter how they like to make it look in the text.

[nancylebov.livejournal.com]

When I saw "horses are closest to carnivora", my first thought was The Book of the New Sun.

[rysmiel.livejournal.com]

This strikes me as no more nor less plausible than any of the many other papers that have come out in the last ten years or so arguing various possibilities for mammalian reordering; my PhD supervisor was interested in the topic sort of in passing, and it's not anything I'd regard as settled yet. It's pushing at the edges of the resolution available with the technique used.

How it stands with fossils really depends on whether you believe in molecular clocks or in morphological differences as more reliable, which really is a matter of faith - there are otherwise sensible people who appear to think that morphological data are holy writ and molecular clocks that disagree are therefore bunk, and vice versa.