Molecular Clocks


Over at Dembski & Co., there are some recent posts (here, here, and here) complaining about molecular clocks and arguing that well known and long established limitations of molecular clocks invalidate evolutionary biology.

I find it sad that a week after scientists used molecular clocks to show that the Tripoli Six did not cause an HIV outbreak, the anti-evolutionists at UD throw up some posts ignorantly questioning the well established and understood procedure. I feel like arguing that anti-evolutionists want these six innocent health workers to be executed, but I’m sure that is not the case. They just don’t get the science or even care to. But the science is important, and the ignorance engendered by anti-evolution can have life-and-death consequences.

I don’t know what the people of UD learned in their evolutionary biology graduate programs, but in my program we were taught that the molecular clock is a methodological assumption and that if we use it long enough, we will work with datasets that violate it.

Now the molecular clock is a rather simple thing. It involves one major assumption: that the rate of substitutions per million years is constant among the molecules (DNA or proteins) that you are comparing. Yes, that is an assumption, and science typically makes assumptions when it tries to estimate or calculate some measurement of interest. For example, chemistry students work with ideal gases, and physics students work with ideal springs. Likewise, biology students work with ideal organisms.

The simple procedure looks like this. First you need some DNA or protein sequences. If you know the time of separation between at least two of your sequences, then you can calculate a rate of substitutions per million years from the pair. Using this rate and the above assumption, the time of separation between any two of your other sequences can be estimated. This can allow one to estimate when speciation occurred or a taxon was formed. Of course, this procedure can be made more complex as needed. There are models that use relaxed molecular clocks, multiple calibration points, etc., and many evolutionary biologists don’t even work with molecular clocks. Thus, strict molecular clocks are not as pervasive as some would think.

Clearly from what we know about organismal evolution, datasets with organisms that are closely related or distantly related can and probably will violate the assumptions of the molecular clock: the former because not enough generations have elapsed for the weak law of strong numbers to average out evolutionary variation and the later because evolution can and will make rates in distantly related species uncorrelated.

Now in my grad program, we were also taught that you must calibrate every molecular clock that you do. You can’t take a molecular clock estimate from plants and apply it to mammals and expect to get reasonable estimates of divergence time. You can’t take an estimate based on mutations in a pedigree and apply it to substitutions between species and get reasonable estimates of divergence time. This is known and not a surprise, yet anti-evolutionists have acted like this is some new nail in the coffin of evolution. (I will point out that I’ve taught such limitations to undergraduate evolutionary biology students. Too bad more people haven’t had me for a TA.)

We know these limitations of molecular clocks. We knew them decades before ID was a twinkle in creationists’s eyes. Does that mean that scientists don’t make mistakes when it comes to applying molecular clocks to real data? Of course not. We can expect that scientists, especially ones lacking a thorough training in molecular evolution, will make mistakes when applying this procedure. We can also expect mistakes to occur if the procedure is applied when there is limited data, as can often be the case when someone is trying to build a dataset on some obscure taxa combining results from different fields like biochemistry, genetics, paleontology, and paleogeology. We can also expect that estimates, derived from the molecular clock assumption, can be very wrong and contradictory. But that can happen when you are estimating something.

The bottom of the story is that the anti-evolutionists are not telling us anything that we don’t already know and don’t already work around. (In fact, they often lift such cautions from papers of evolutionary biologists without mentioning to their congregations that evolutionists are pointing these things out, not creationists.)

Ideally, I’d include several references in this post, but I’m nine hours from campus and when I get back I’ll be continuing my work on analyzing human, chimp, rat, and mouse genomes using non-molecular clock models. (I guess that the Dogmatic Darwinists haven’t gotten to me yet.) Hopefully, one of the other pandits that works with molecular data will fill in the blanks this week for our readers.


Ideally, I’d include several references in this post, but I’m nine hours from campus

There you go.

The guy doing some of best stuff with relaxed molecular clocks (and without clocks too) Michael Sanderson and you can get most of his publications from that page. I recommend these papers:

Near, T. J., and M. J. Sanderson. 2004. Assessing the quality of molecular divergence time estimates by fossil calibrations and fossil-based model selection. Phil. Trans. R. Soc. London 359:1477-1483.

Sanderson, M. J. 2003. r8s; inferring absolute rates of evolution and divergence times in the absence of a molecular clock. Bioinformatics 19:301-302.

The most fundamental property of molecular clocks is that they are monotonic; they never run backwards. From the standpoint of evolution “mutation time” is at least as significant as chronological time.

Even if the rates of molecular clocks have large variances (due, say, to hundreds of reversals of the Earth’s magnetic field which then subjects life to large periodic increases in extraterrestrial radiation), molecular clocks are still useful.

No reputable scientist would use a single clock without cross checks (which the anti-evolutionists often imply in their critiques of the methods of science). Multiple methods are used in science as a matter of routine. Anti-evolutionists either don’t know this, or they deliberately misrepresent scientific methods.

I happened to look at the ReasonsToBelieve the other day. One thing I found that was rather funny was the fact that they had an article arguing that molecular clocks confirm the Biblical timeline (actually, it was off by a factor of 10, but they claimed it confirmed the Biblical timeline). The funny part was the fact that two articles earlier they attacked molecular clocks and claimed that they were effectively useless. So which is it? Are they useful when they confirm the Biblical timeline, and otherwise completely useless?

The most fundamental property of molecular clocks is that they are monotonic; they never run backwards.

Whereas the most fundamental property of ID speculations is that they are monotonous: they never move forwards.

Don’t miss DaveScot’s “enlightened” response to the above post.

Reed A. Cartwright linked:

… DaveScot’s “enlightened” response to the above post.

This struck me as funny, on UD tribune7 wrote:

Exactly! We are the Copernicans. :-)

Ummm… weren’t the Copernicans able to calculate planet positions more accurately, and then only after they figured out eliptical orbits.

What exactly can the IDiots calculate?

Could their ID model somehow prove the Tripoli Six innocent or guilty?

Could their model tell us where to find fossils, or predict what kind of fossils we will find?

Of course they’re the Copernicans! (Just to double-check, Copernicans believe that God moves the planets, right? And the pre-Copernicans believe that they can use mathematics to calculate planetary movements?)

Dear God, these people are such clowns. First Dembski with that sad and pathetic video, and now this intellectual drool. If molecular clocks are so bad, perhaps they can explain the success of molecular clocks in predicting events in the fossil record which have subsequently been corroborated by real evidence? My field – physical anthropology – was revolutionized when molecular clocks were used to predict that chimps and humans shared a common ancestor only about 7 million years ago, give or take a few. The prevailing view at the time was that the common ancestor would be found about 15 million years ago. I remember taking a class from Elwyn Simons (one the great paleontologal experts on primate evolution) wherein he was positively ranting about how wrong these clocks must be. Sadly, the paleontological opinion was based on gut feeling, and had taken on too much inertia. Now, after 25 years of field work, all indiciations are that the common ancestor should be around 7 million years, give or take a bit. Isn’t that just like science? You make a prediction and test it and confirm it with independent data! It’s especially good when the prediction runs so contrary to the standard view. But of course, we don’t hear about this stuff from the “scholars” of the creationist camp. No, their brand of “scholarship” is to crap in their hands and throw it at scientists.

Now to make another prediction. My guess is that the standard creationist response will be to claim that the human fossil record is being interpreted to force the findings to conform to the molecular clock. Wrong. The paleontologists were FORCED to accept the clock results by virtue of the data that were (and are being) collected from the field. Were this a matter of interpretation, the opposite result would have been obtained.

They claim that molecular clocks are the equivalent of epicycles???? Good heavens, that’s idiotic.

There are no additional mechanisms (little clocks ticking?) that need to be assumed along with molecular clocks. Are they questioning the fact that DNA sequences mutate, or that their mutation frequencies may be different? The only question is whether changes accumulate at relatively constant rates (or ‘constant’ enough for us to predict something based on them).

In contrast, an explanation comparable with epicycles would be to tack on an unnecessarily complicated mechanism which is responsible for the observed changes in DNA. Like, say, AN INTELLIGENT DESIGNER.

What seems to be overlooked is that the molecular clock hypothesis was extremely controversial from the very beginning and that there has never been a consensus that it can work. To be sure, it has made some spectacular predictions (ie human-chimp divergence time) but it has also resulted in some predictions that are very difficult to accept.

The critical commentary against the molecular clock has been continuous to this day (as it should be). The paper they cite is just the latest in a long line of criticisms. The debate occurs at a very technical level but if these technical problems are not resolved it could just turn out that using the molecular clocks to date very old divergences (ie > 100 MYA) may not be possible. I would hate to see that occur, but the theory of evolution wouldnt fall because of it, just that we would be deprived of a potentially important tool

Reading the jokers over at the DI thread shows that they are completely ignorant of the science in molecular evolution. They confuse the molecular clock hypothesis with the notion of constant molecular clocks and seem to unaware that phylogenies can be deduced independently of any rate assumption. Some really bizzarre comments: ie “Darwinism requires molecular clocks”?? How can they possilby detect devine intervention in such technical difficulties?

Sal’s comments to me over at UD seem irrelevant in my opinion. Last time I looked evolutionary biologists did not think that there was some universal clock for substitutions, nor do they think that all evolutionary patterns are explained by clock-like processes. It seems that Sal is trying to have an argument that doesn’t exist.

In my opinion, there is no molecular clock at work in molecular evolution. Substitution is a stochastic process, not a deterministic, clock-like one. However, this does not mean that under certain conditions that a deterministic process is not a good approximation for a stochastic one. (Physics is filled with them.) That’s all a molecular clock is, an approximation for the complexities of real life. It is a tool, not some underlying fundamental process or principal.

Anyone familiar with the math will realize that we employ the molecular clock assumption to get around limitations in maximum-likelihood (frequentist) approaches to estimating species divergence. That is why we have a grant in my lab that explores Bayesian techniques that can avoid assuming molecular clocks.

Reed, your statement “Substitution is a stochastic process, not a deterministic, clock-like one” has something wrong with it. If substitution were a time dependent stochastic process than you could very well make a clock out of it provided that the process could be modeled by a Poisson distribution. One could make a very good clock out of radioactive decay for example.

The problem with finding a molecular clock is finding distribution function that is itself is not time dependent.

Mike, whether a process can be approximated by a clock is different than whether it actually is clock-like. The distinction is whether X happens exactly every four seconds or whether X happens on average every four seconds. The former can be similar to the later depending on the distribution of the later and the length of time you are looking at.

You are correct that Poisson processes can lead to rather easy clock approximations, but molecular clock approximations can also be calculated for other distributions as well.

About this Entry

This page contains a single entry by Reed A. Cartwright published on December 17, 2006 1:12 AM.

Dembski as the David Seville of Information Theory was the previous entry in this blog.

Dembski’s motive is the next entry in this blog.

Find recent content on the main index or look in the archives to find all content.



Author Archives

Powered by Movable Type 4.361

Site Meter