Indel Length Distributions and Evolutionary Divergence

Posted 6 December 2008 by Reed A. Cartwright

Some of you may remember that several years ago that Britten (2002) argued that human-chimp divergence was 5% not ~1.2%. (See this press release for a refresher.) Of course, creationists jumped on this research and began harping that the more scientists looked, the more distant humans and chimps were. This is important to them because the number one rule of creationism is "no matter what, humans are not related to any other living creatures," which is so difficult to maintain in our age of science and education.---Amusingly, humans and chimps are so similar to one another that creationists cannot create a consistent definition of "created kinds" that makes humans special and lumps all the boring animals together. Britten (2002) derived his 5% divergence metric by considering the lengths of insertions and deletions (indels) along with point substitutions between human and chimp genomes. This is unlike other estimates that just consider the number of point substitutions that have occurred between the two species and find ~1.2% divergence. At the time I commented that these two numbers---1.2% and 5%---could not be compared because they are different metrics. Additionally, Britten's metric is probably unfairly upweighting the contribution of indels because a single event can add or remove multiple residues at a time. A recent study of mine, which was not directed at Bitten's work, has found that it is actually worse than that. Simply put, the total length of indels separating humans and chimps is unrelated to the evolutionary divergence between them. This arises because the variance of indel length is "nearly-infinite", which causes nonconservation of average indel length. Therefore, two pairs of species, equally divergent evolutionarily, can and probably will have very different proportions of nucleotides belonging to indels. One pair might be 5% divergent and the other 1.5% divergent, including indels, without any underlying change in the evolutionary process or time since speciation. The upside is that traditional substitution based evolutionary distances are unaffected and can still be used to properly estimate the evolutionary divergence between species.

Britten, R. J. 2002. Divergence between samples of chimpanzee and human DNA sequences is 5%, counting indels. Proc. Natl. Acad. Sci. USA 99:13633--13635. [link]

Cartwright, R. A. 2008. Problems and solutions for estimating indel rates and length distributions. Mol. Biol. Evol. Advance Access. [link]

29 Comments

iml8 · 6 December 2008

Not picking on this item since I found it interesting
but it would seem from what you say that the "percent difference metric" depends,
as is often the case with metrics, on the specific
assumptions used to define it -- different assumptions
will give somewhat different metrics, to get consistent
you have to get a consensus on the assumptions. Not like,
say, measuring speed in a race where the metric is fairly unambiguous. Right?

Is the notion of the humans as the "third chimp", with
chimps, humans, and bonobos rougly equidistant, still
regarded as valid? From the Darwin-basher point of view,
the issue is nitpicking, since it's clear that humans
are more closely related to chimps than any other primate,
and even if that is rejected, humans are clearly more
closely related to the great apes than they are to gibbons, much less monkeys and lemurs.

White Rabbit (Greg Goebel) http://www.vectorsite.net/gblog.hml

DS · 6 December 2008

Thanks for the link Reed.

Reed A. Cartwright · 6 December 2008

iml8 said: different assumptions will give somewhat different metrics, to get consistent you have to get a consensus on the assumptions. Not like, say, measuring speed in a race where the metric is fairly unambiguous. Right?

There is a consensus metric: substitution time---expected number of substitutions per site. Different models of molecular evolution will estimate different substitution times but Britten's metric was very different than the standard metrics. According to the results of my study, Britten's metric is probably a poor choice for measuring evolutionary divergence. But luckily, it doesn't seem to be used that much.

iml8 · 6 December 2008

Reed A. Cartwright said: There is a consensus metric: substitution time---expected number of substitutions per site. Different models of molecular evolution will estimate different substitution times but Britten's metric was very different than the standard metrics.

Ah, good clarification, thanks. Does anyone know what the current status of the "third chimpanzee" notion is? I was under the impression it was a bit controversial. Well, more than a bit for Darwin-bashers. White Rabbit (Greg Goebel) http://www.vectorsite.net/gblog.html

Reed A. Cartwright · 6 December 2008

iml8 said: Does anyone know what the current status of the "third chimpanzee" notion is? I was under the impression it was a bit controversial. Well, more than a bit for Darwin-bashers.

I don't hear anyone talking about it out side of Diamond's book. It's a catchy phrase, but nothing much beyond it. First, chimps and bonobos are more related to each other than they are to humans, and humans are equally related to chimps and bonobos. The actual relationships look like this: ((chimps,bonobos),humans); Second, according to the rules of taxonomy if Pan and Homo genera were collapsed, the new genus would be named Homo because it is the older name, making them "human" not us chimps.

Frank J · 6 December 2008

This is important to them because the number one rule of creationism is “no matter what, humans are not related to any other living creatures,”
— Reed A. Cartwright

If the convention is to include ID under the "creationism" banner, then I wouldn't call that the "number one rule." Of course that doesn't stop IDers, even those who admit common descent, from giving their audience another excuse to deny it.

Reed A. Cartwright · 6 December 2008

Most IDists do not accept that humans are related to other species. At the Kansas Kangaroo Kourt a few years ago only a handful of the ID witnesses said that they accepted human and chimp common descent. The vast majority rejected it. Behe might be a well know acceptor of human-chimp common descent, but much of his allies out right reject it.

Mark Frank · 7 December 2008

Is it a coincidence that this interesting and informed article was posted on the same day this on Uncommon Descent (also interesting in a way)?

Reed A. Cartwright · 7 December 2008

I don't bother with the rantings at UD. It's just a coincidence.

Frank J · 7 December 2008

Behe might be a well know acceptor of human-chimp common descent, but much of his allies out right reject it.
— Reed A. Cartwright

Sure, and as you know, yet other IDers don't commit either way. What I mean is that, while classic YECs and OECs obsess over humans vs. other species, and can't even bear the thought of a possible in-vivo "saltation" event, IDers prefer to avoid the subject in favor of more remote "blessed events" such as the first bacterial flagellum or the Cambrian "explosion." Given the makeup of its audience, however, an argument can be made that by avoiding the subject ID actually raises its visibility.

Frank J · 7 December 2008

Reed A. Cartwright said: I don't bother with the rantings at UD. It's just a coincidence.

Did you run it through the EF? ;-)

gofor · 7 December 2008

When a creationist uses a word, they don’t mean anything remotely similar to the actual definition. They’re just twisting the language into whatever knots they need to to promote their lies. You will never, ever, ever get a creationist to use the real definitions of words.

Can anyone give an example of where creationists 'twist the language' etc??

Stanton · 7 December 2008

gofor said:
When a creationist uses a word, they don’t mean anything remotely similar to the actual definition. They’re just twisting the language into whatever knots they need to to promote their lies. You will never, ever, ever get a creationist to use the real definitions of words.
Can anyone give an example of where creationists 'twist the language' etc??

When creationists use the word "theory" to dismiss evolutionary biology.

Cheryl Shepherd-Adams · 7 December 2008

gofor said: Can anyone give an example of where creationists 'twist the language' etc??

Many examples here.

Stanton · 7 December 2008

gofor said:
When a creationist uses a word, they don’t mean anything remotely similar to the actual definition. They’re just twisting the language into whatever knots they need to to promote their lies. You will never, ever, ever get a creationist to use the real definitions of words.
Can anyone give an example of where creationists 'twist the language' etc??

Also, how is this, a quote from another thread, directly relevant to this specific thread?

Thomas · 7 December 2008

I continue to find it fascinating that creationists even try to examine legitimate scientific evidence. This only underscores how weak their axioms are.

If, hypothetically, a fully modernized force like the US Marine Corps with firearms, aircraft, landmines and modern tactics were to fight an open field battle with a stone age tribe armed with stone hand axes and clubs carved from the bones of mega-fauna the Marines, even if vastly outnumbered, probably wouldn't spend much time contemplating the strategy of their opponents. They would simply annihilate them and be done with it.

Creationists, who fashion themselves the ideological Marines in this analogy, seem hugely interested in the flint knapping and bone carving techniques of the tribe they think that they're fighting. Despite creations claim of a monopoly on truth they obviously believe they have no such thing.

iml8 · 7 December 2008

Reed A. Cartwright said: First, chimps and bonobos are more related to each other than they are to humans, and humans are equally related to chimps and bonobos. The actual relationships look like this: ((chimps,bonobos),humans);

Yeah, that makes sense. I know physical appearance isn't necessarily correlated to genetic difference -- wolves versus pekinese for example -- but I think even if an alien observed the three species, the intuition would be to lump bonobos and chimps together, then link in humans. Jared Diamond is a clever guy but I think sometimes a bit too clever for his own good. Somewhat verbose, too -- I tried to read through GUNS GERMS & STEEL despite my impression that it would be a good book at half, even a quarter the length. It was. I must admit that the chapters on the coevolution of humans and their domesticated organisms were extremely interesting, however. I'll have to outline them for my blog. White Rabbit (Greg Goebel) http://www.vectorsite.net/gblog.html

Dave C · 7 December 2008

Regarding the original claim by Britten, I never found it to be all that controversial, at least from an evolution vs. creationism perspective. Isn't it correct to say that if Britten were right about human-chimp divergence being more like 5%, we would also have to increase the percentage for, say, human-elephant divergence, human-cantaloupe divergence, etc.? If so, that would just increase the divergence across the board, but it wouldn't really call into question the notion that humans are most closely related to chimps/bonobos. Am I right on this?

John Kwok · 7 December 2008

Frank, You forgot to ask Reed if he had calculated 95% confidence limits, which is what I have asked of Dembski both in person and via e-mail:

Frank J said:
Reed A. Cartwright said: I don't bother with the rantings at UD. It's just a coincidence.
Did you run it through the EF? ;-)

Appreciatively yours, "Kohn" Kwok

slpage · 7 December 2008

"Regarding the original claim by Britten, I never found it to be all that controversial, at least from an evolution vs. creationism perspective. Isn’t it correct to say that if Britten were right about human-chimp divergence being more like 5%, we would also have to increase the percentage for, say, human-elephant divergence, human-cantaloupe divergence, etc.? If so, that would just increase the divergence across the board, but it wouldn’t really call into question the notion that humans are most closely related to chimps/bonobos. Am I right on this?"

Very much so.

Of cours,e creationists never seem to want to talk about that.

Reed A. Cartwright · 7 December 2008

John Kwok said: You forgot to ask Reed if he had calculated 95% confidence limits, which is what I have asked of Dembski both in person and via e-mail:

The Fisher information matrix of the maximum likelihood estimates was calculated and inverted to get a covariance matrix. See Figure 3.

Stanton · 7 December 2008

slpage said: "Regarding the original claim by Britten, I never found it to be all that controversial, at least from an evolution vs. creationism perspective. Isn’t it correct to say that if Britten were right about human-chimp divergence being more like 5%, we would also have to increase the percentage for, say, human-elephant divergence, human-cantaloupe divergence, etc.? If so, that would just increase the divergence across the board, but it wouldn’t really call into question the notion that humans are most closely related to chimps/bonobos. Am I right on this?" Very much so. Of course, creationists never seem to want to talk about that.

Of course, creationists never seem to want to talk about anything scientific to begin with.

John Kwok · 7 December 2008

Dear Reed, At least you can calculate 95% confidence limits:

Reed A. Cartwright said:
John Kwok said: You forgot to ask Reed if he had calculated 95% confidence limits, which is what I have asked of Dembski both in person and via e-mail:
The Fisher information matrix of the maximum likelihood estimates was calculated and inverted to get a covariance matrix. See Figure 3.

A major statistical problem with the EF is that you can't calculate confidence limits, and that really demonstrates IMHO what an abysmal theoretical statistician Dembski claims to be. There's simply no way you can recognize potential Type I and Type II errors. Thanks, John P. S. You may have also noticed that I had a slight "dig" for Frank too.

JimF · 7 December 2008

Reed said: This is important to them [creationists] because the number one rule of creationism is "no matter what, humans are not related to any other living creatures," which is so difficult to maintain in our age of science and education.

Which is why the Dmanisi fossils are so wonderful - creationists are clueless about how to classify them. See my recent posts: Dmanisi and Answers in Genesis, and
Dmanisi fossils - more transitional than ever

Henry J · 7 December 2008

iml8: Yeah, that makes sense. I know physical appearance isn’t necessarily correlated to genetic difference – wolves versus pekinese for example – but I think even if an alien observed the three species, the intuition would be to lump bonobos and chimps together, then link in humans.

Intuition would also link lizards and crocodiles together, apart from birds.

Dave C: Isn’t it correct to say that if Britten were right about human-chimp divergence being more like 5%, we would also have to increase the percentage for, say, human-elephant divergence, human-cantaloupe divergence, etc.?

I'm not sure. One would have to be careful about including counts of types of change that might occur in large numbers in one species, since that situation could make the one species look like the outlying member of the group when it isn't. Henry

Paul Burnett · 7 December 2008

Henry J said: Intuition would also link lizards and crocodiles together, apart from birds.

While lizards and crocodiles are more obviously related, I would defy you to look at some lizards' lower legs (tib/fib) and feet and some bird's lower legs/feet in isolation and tell which was which - they're amazingly similar.

Henry J · 7 December 2008

While lizards and crocodiles are more obviously related, I would defy you to look at some lizards’ lower legs (tib/fib) and feet and some bird’s lower legs/feet in isolation and tell which was which - they’re amazingly similar.

Ah, but that by itself would put birds closer to lizards than to crocodiles (they aren't). Otoh, intuition does depend on how much detail the person observes before applying the intuition. Henry

Stephen Wells · 8 December 2008

On metrics:

Imagine a book in ten volumes.
Each volume has 10 pages.
Each page has 10 lines.
Each line has 10 words.

Change one word.

10% of volumes have changed.
1% of pages have changed.
0.1% of lines have changed.
0.01% of words have changed.

Now it turns out that counting duplicated lines isn't helpful.

Mike from Ottawa · 8 December 2008

And 100% of your sample of ten volume sets of books has changed.