A supertree showing mammalian evolution

March 28, 2007 3:27 PM

The significance of the dinosaurs' death has been greatly exaggerated. This article in Nature discusses how mammalian evolution accelerated independent from the death of dinosaurs. The theory was derived from a "supertree" [pdf ~ 1mb] of mammals and how common ancestors have branched out. Coolest info-graphic ever.
posted by phyrewerx (33 comments total)

The article from Nature.
The author of the pdf.
Check out the pdf, you won't regret it.
posted by phyrewerx at 3:30 PM on March 28, 2007

I can regret anything!
posted by nathancaswell at 3:36 PM on March 28, 2007

A huge PDF is really not a good interface for this kind of radial tree display.

How about some focus + context zooming?

I do like the "You are here." Nice touch.
posted by demiurge at 3:36 PM on March 28, 2007

Sweet pdf. I would like a poster of that. It looks like death star plans when zoomed out.
posted by Totally Zanzibarin' Ya at 3:44 PM on March 28, 2007

I don't know much about it, but who postulated that the fall of the dinosaurs accelerated the evolution of mammals? I was never taught that.

Yeah, that PDF is screaming for a Flash version.
posted by mrgrimm at 3:49 PM on March 28, 2007

I haven't read the paper yet (I'm in a coffee shop, so don't have my university access to Nature online), but I think it was long held that the rise of mammals was made possible by the extinction of the dinosaurs at the K/T mass extinction.
posted by HighTechUnderpants at 3:56 PM on March 28, 2007

Supertree? Perfectly average tree, I say. It has no cape.

Neat stuff.

mrgrimm: I don't know much about it, but who postulated that the fall of the dinosaurs accelerated the evolution of mammals? I was never taught that.

I was under the impression it had to do with the opening-up of a massive amount of niches. Don't know if that qualifies as "accelerating" their evolution.
posted by CKmtl at 3:59 PM on March 28, 2007

mrgrimm: but who postulated that the fall of the dinosaurs accelerated the evolution of mammals? I was never taught that.

Someone must have put forth this idea long ago, and these days it is popular folk science. For example this story in the Boston Globe says:

The dinosaurs died out, making room for mammals and eventually for man.

posted by MonkeySaltedNuts at 4:10 PM on March 28, 2007

PDF nicely confirms a rather humbling statistic I heard years ago-- we mammals are about 40% rodent.
posted by jamjam at 4:39 PM on March 28, 2007

Just so no one is confused, that kind of graphic is not called a supertree. A supertree is what the author calls a tree created by the combination of multiple trees.
posted by grouse at 5:13 PM on March 28, 2007

jamjam: we mammals are about 40% rodent.

and about 20% bats. And humans' closest major group relatives are rodents and rabbits, not something noble like lions, horses, or elephants.
posted by MonkeySaltedNuts at 5:15 PM on March 28, 2007

I just took a Principles of Evolution class. We were taught the principle that contrary to what you'd expect, species aren't driven extinct (usually) by competition from other species invading their adaptive role. Rather, it's most often by random environmental factors (luck of the draw, environmental change, what have you) which leaves their adaptive role open for other lineages to adapt into.
Which means that though according to this study all the major lineages of mammals had been established before the K/T extinction, dinosaurs still had to die out to leave open roles for mammals to ease into, such as most megafauna roles which are currently filled. The mammals may have been doing well under the rule of the dinos, but don't assume that they outcompeted them to the point of extinction.
posted by lostburner at 5:30 PM on March 28, 2007

And humans' closest major group relatives are rodents and rabbits, not something noble like lions, horses, or elephants.

In the most commonly used sense of "close," that is correct—the last common ancestor between humans and mice occurred much more recently than, say, the LCA between humans and dogs. But those rodents reproduce so darn fast that more mutations have accumulated in the rodent lineage. Because of this, a human genome actually has more in common with a dog genome than a mouse genome.
posted by grouse at 5:32 PM on March 28, 2007

Kinda ruins that Simpsons evolution post from the other day, don't it?
posted by evilcolonel at 5:43 PM on March 28, 2007

I had to blow up the PDF by 2400% to read the species names, which translates to a 22'x17' poster if printed out in hard copy.

That's both really annoying and cool as hell.
posted by LordSludge at 5:55 PM on March 28, 2007

I just finished reading the nature article this was from. All of the dates in the supertree were generated using a molecular clock. Molecular clocks are notorious for overestimating dates. This study used 30 fossil calibration points (which is a lot better than most molecular clock studies actually), but took them to be minimum divergence time estimates. This inherently biases the results towards creating dates that are deeper in time than any fossil evidence. It basically comes down to the more you distrust the fossil record the more your results are not in keeping with it. I therefore trust these results about as much as the authors of the study trust the fossil record. Which is to say that I don't.

That said, I am a grad student in paleontology with a strong inherent bias towards trusting the fossil record so feel free to not trust anything I say on the matter.
posted by DanielDManiel at 6:08 PM on March 28, 2007 [2 favorites]

Actually, ec, I think The Simpsons might agree. As you can tell, Homer's descendant does not really mess with the Dinosaurs. Other than possibly the nice competitive hurdle of a bigger, stronger predator being taken away rather conveniently, the theory has a point.
posted by skepticallypleased at 6:19 PM on March 28, 2007

Coolest info-graphic ever.

Beg to differ. Truly terrible information design, doesn't really work at any zoom level (if you're close enough to read, you lose the structure, if you're far enough to see structure, can't read names), and the radial layout has nothing to do with the information it contains, and distracts from the actual relationships, that would be much better shown as a traditional orthogonal tree.
posted by signal at 6:28 PM on March 28, 2007 [1 favorite]

What signal said. I can read those tiny letters only if I blow it up to 1600%.

Also, what about the megafart 55 million years ago? That changed the climate and paved the way for us and other mammals, sez Fred Pearce. Never mind the 65 mya event.
posted by Listener at 6:43 PM on March 28, 2007

This study used 30 fossil calibration points (which is a lot better than most molecular clock studies actually), but took them to be minimum divergence time estimates. This inherently biases the results towards creating dates that are deeper in time than any fossil evidence.

You write as if there is a choice other than taking fossil dates as minimum estimates. If you can accurately date a fossil, the only thing you know for certain about when the fossil's species diverged is that it happened before the date of your fossil.

It basically comes down to the more you distrust the fossil record the more your results are not in keeping with it. I therefore trust these results about as much as the authors of the study trust the fossil record. Which is to say that I don't.

It's ridiculous to say that the authors do not trust the fossil record, because they would not be able to provide times at all without using it (as they did). So why didn't they use more fossil calibration points? It would be impossible to include every known fossil date because they aren't in a central database that can be searched automatically (such as the 534,225 sequences from 3,179 species used in this computation). Would it have been better if they had used more? Yes. But they had to stop somewhere and I think their decision to take a fossil from each "major lineage" was sound.

If you paleontologists want us evolutionary biologists to use more of your data in this sort of thing, then start putting it in a free database that can be easily mined, as we have been putting every bit of our sequence data the last 25 years. We will gladly use it.

Truly terrible information design, doesn't really work at any zoom level (if you're close enough to read, you lose the structure, if you're far enough to see structure, can't read names), and the radial layout has nothing to do with the information it contains, and distracts from the actual relationships, that would be much better shown as a traditional orthogonal tree.

I doubt it was "designed" to be viewed as a PDF, despite the BBC's enthusiasm in including the full tree. I've seen trees like this as enormous posters before. The figure published in Nature is much better for viewing on a computer monitor.
posted by grouse at 7:29 PM on March 28, 2007 [2 favorites]

Damn, that's crying out for a wide-format printer. A good 60 inches if not more.
posted by five fresh fish at 7:40 PM on March 28, 2007

Mammals kick ass.
posted by neuron at 7:40 PM on March 28, 2007

Who will kick their (our) ass?
posted by rosswald at 8:48 PM on March 28, 2007

The actual paper has a much reduced version of the pdf, leaving out all the terminal details. If you have access it is here.

The program that drew the tree is free, and here, and allows all sorts of manipulations, but I cant find the raw data for the tree.
posted by scodger at 10:42 PM on March 28, 2007

Who will kick their (our) ass?

Cockroaches and ocean jelly.
posted by five fresh fish at 11:14 PM on March 28, 2007

That supertree would be a lot more informative for dummies like me if they used the common names instead of latin names. "Elephantodinea" I can grok. Many of the others are beyond my kin.
posted by five fresh fish at 11:17 PM on March 28, 2007

You write as if there is a choice other than taking fossil dates as minimum estimates.

Good point, obviously any first possible appearance of a taxon is a minimum constraint on the true origination of the taxon, but the way the authors treat "minimum divergence times" is -as I understand it from the paper's supplemental material- to throw out calibration points that seem too young based on what the rest of the calibrations show and replace them with dates derived from the rest of the calibration points. This seems reasonable, but if you don't additionally throw seemingly too old calibration points -which can be caused by poor taxonomy or bad chronostatigraphic dates- then you are biasing your results towards old divergence times. I think using biostratigraphic uncertainty intervals which come from the spottiness of finds would be a better way to deal with minimum constraint part of the problem albeit way more difficult.

If you paleontologists want us evolutionary biologists to use more of your data in this sort of thing, then start putting it in a free database that can be easily mined, as we have been putting every bit of our sequence data the last 25 years

I wholeheartedly agree that there should be more collaboration between paleontologists and evolutionary biologists and that all data should be more accessible, but as it stands I feel like the paleontological data that is used for molecular clock work is mishandled. Molecular clock studies are still relatively new and results often disagree. They also seem to be hot shit at the moment and I just feel that too many people are jumping on the bandwagon and trusting the results without questioning them. Then they get presented to the public as hard fact requiring an accommodating paradigm shift in paleontology. If you need to trust the fossil record at all for calibration, then really get to know it or work with someone who already does.
posted by DanielDManiel at 2:28 AM on March 29, 2007

That PDF is horrible. That said, here's a similar one showing the family tree of plants, animals, fungi, and protists. I saw it at the Museum of Contemporary Art in Chicago at the Future of Global Design exhibit where it was presented quite compactly on an illuminated table. Slightly easier, but still hard to read that way.
posted by dgaicun at 4:53 AM on March 29, 2007

Anyone following this closely might be interested in the 1998 Nature fossil debate. Although I haven't read it.

Sorry about the multiple comments coming up—MeFi isn't accepting a single big post.
posted by grouse at 7:49 AM on March 29, 2007

Many of the others are beyond my kin.
posted by five fresh fish

Eponysterical!
posted by grouse at 7:50 AM on March 29, 2007

...the way the authors treat "minimum divergence times" is... to throw out calibration points that seem too young based on what the rest of the calibrations show and replace them with dates derived from the rest of the calibration points. This seems reasonable, but if you don't additionally throw seemingly too old calibration points -which can be caused by poor taxonomy or bad chronostatigraphic dates- then you are biasing your results towards old divergence times.

I see what you are getting at here, and take your point. As far as I see it, there are two kinds of error inherent in taking the oldest known fossil dates as the actual divergence dates:

1. error due to the simple fact that the divergence always must have occurred before the date of the fossil
2. error due to problems with the date measurements

While the first kind of error must be systematic, and always in the same direction, I would have assumed that the latter kind of error would be random. Therefore it would be appropriate to have a systematic correction for the first type of error, but not the second, as for every fossil date that is observed older than the true date, you are likely to have a fossil date that is observed younger than the true date. Please correct me if I am wrong on this. But if I am wrong, and the fossil dates themselves, not just divergence dates, are systematically overrestimated by paleontologists, then one can hardly be blamed for seeking an alternative way of estimating these dates. (Of course it doesn't help the big conclusion of this study though.)

One must also consider the conservative estimate of 166 Mya as the root (entirely based on fossil dates), which would increase bias in this study towards shorter timescales.

I think using biostratigraphic uncertainty intervals which come from the spottiness of finds would be a better way to deal with minimum constraint part of the problem albeit way more difficult.

You are right that it would have been better to include stratolikelihood analysis in this work, just as it would have been better to include more fossil calibration dates. It also would have been way more difficult.

Paleontologists could already have produced a tree of this magnitude based on stratolikelihood (and not just assumption of divergence time = stratigraphic time), without the need for recourse to molecular data. I would be curious to know if they have they done so.

I feel like the paleontological data that is used for molecular clock work is mishandled.

You are definitely right, at least some of the time. Many molecular clock studies in the past were extremely flawed, especially by ignoring the fact that a stratigraphic date has inherent random error, or using only a couple of calibration points. Thankfully this paper does not suffer from those particular problems.

Molecular clock studies are still relatively new and results often disagree.

They've been around for more than 40 years which is considered ages in molecular biology. Perhaps not so much in paleontology. Results do often disagree, you are right.

If you need to trust the fossil record at all for calibration, then really get to know it or work with someone who already does.

I'll note that there's at least one paleontologist author, who I expect knows the fossil record pretty well.

//

On reflection it is interesting to consider the implications of this work on our respective fields, and how this colors our own reaction to this work. You might see the article's challenge to paleontological dogma as an extraordinary claim that is not supported by a single flawed study. I, on the other hand, am less interested in that claim either way, and am more interested in having a detailed tree of the mammals that appears to be better than what has been previously available.
posted by grouse at 7:51 AM on March 29, 2007

I love the graphic. My personal opinion is that Acrobat is not the best tool (why can't I zoom and pan with one single tool like google maps or local.live.com?). However, I suspect if I had a huge monitor, say, the size of a wall, this graphic would be navigable.
posted by jeffamaphone at 9:39 AM on March 29, 2007

This is my last comment as I feel sufficiently out argued, so here's the watered down version of what I have to say on the subject.

1) Molecular clock work, like the paper being discussed, still creates divergences times that may be systematically biased. Although, mostly what I have read before had to do with the last common ancestor of all bilaterians and that work was in such disagreement with itself and the fossil record and used so few fossils that it made my blood boil.

2) Phylogenetic trees like the one posted are great and wonderful for everybody. I just hope the dates get better. By better I mean free from systematic error and not just more in keeping with fossil evidence. I am perfectly happy to eat my hat, just not quite yet.

3) I for one like the pretty circular diagram, and don’t have any trouble extracting information from it.
posted by DanielDManiel at 5:03 PM on March 29, 2007