Tuesday, October 10, 2006

Review of "The Mystery of the Genome" (II)

At the end of the previous installment I began examining Sanford's arguments as to why "random mutations are never good". As we saw, most of these ended up being among some of oldest and most discredited creationist arguments around. But he also had a new (at least to me) and more subtle one: that even if there are beneficial mutations they'll turn out to be nearly-neutral. Although it's nice to see that creationists have caught up with Kimura, albeit several decades late, I'm afraid this argument is also invalid.

The "nearly neutral" theory of molecular evolution was developed by Motoo Kimura and Tomoko Ohta in the 1970s. Basically this relies on an old result by Kimura that requires an explanation. Imagine that we have a diploid population fixed for an allele A at the A locus. What is the probability that a new allele a arising by mutation (A -> a) will go to fixation some time in the future? The answer is that it depends on whether selection is acting on it. To make a long story short, there are three (simple) possibilities:
  • If the a allele is neutral, that is, the fitnesses of the three genotypes are the same W(AA) = W(Aa) = W(aa), then the probability of fixation of a is P = 1/(2N), where N is the size of the population (number of individuals). In other words, the probability that the a allele will go to fixation due to stochastic effects (or genetic drift) is higher in a small population than in a large one.

  • If the a allele is beneficial, then the probability that it will go to fixation depends on the selective advantage it confers. If the fitnesses of each genotype are W(AA) = 1, W(Aa) = 1+s, and W(aa) = 1+2s, then the probability that a will go to fixation is approximately P = 2s. Other fitness functions will change the probabilities, but that doesn't matter for the main point I want to make.

  • If the a allele is deleterious, then the probability that it will go to fixation depends on the selective disadvantage it confers. The formulas are more complicated and need not detain us for the main point I want to make.
What Kimura and others have pointed out is that if the population size N and/or the selective advantage s of an allele are too small, then a beneficial allele (i.e., one with a positive value of s in the above expressions) will behave as if it were neutral. Such an effectively neutral allele will not be under the action of natural selection, but will fluctuate under genetic drift. A quick manipulation of the probabilities given above shows that if the selective advantage s of a beneficial allele is equal to or less than 1/(4N), then the allele is effectively neutral. A similar point can be made for a weakly deleterious allele.

Sanford's main point then is that there are likely to be very few truly beneficial mutations in humans because of this effect. Although this decades-old argument is generally correct for humans, it is misleading for two reasons. First, it neglects to mention that this near-neutrality effect is especially acute in large mammals like humans because of their historically low population sizes. However, most creatures on earth are not subject to this problem to anything near the same extent. How do we know this? There are many lines of evidence, such as the evolution of codon-usage bias. I won't take the time to explain this in full here, but suffice it to say that it can only evolve if selection is able to act on very weakly beneficial mutations. Briefly, we expect to see strong codon bias in species that have large populations. Predictably we find little or no codon bias in humans or mice (in concordance with Sanford's point), but it is present in nematodes (Caenorhabditis elegans and, more strongly, in C. briggsae), cress Arabidopsis thaliana, fruitflies Drosophila melanogaster, and is very strong in microorganisms like E. coli and yeast Saccharomyces cereviseae. Second, Sanford handwaves about the ratio of beneficial mutations to deleterious mutations, when in fact there are no good direct estimates of this number for humans. Direct estimates in other organisms are not abundant either, because it is technically difficult to do so, but there are some for which the picture is not as apocalyptic as Sanford suggests. For example, Sanjuan, Moya & Elena (2004) found that "the proportion of beneficial mutations was unexpectedly high" in the vesicular stomatitis virus. I also know of at least one other study (as yet unpublished, so I cannot say anything else about it) which found that the ratio of beneficial to deleterious mutations in a famous microbe is much higher than previously thought.

Which brings us to the main problem with Sanford's argument. Let's imagine that what he has said in Chapter 2 is right and that there is no evidence for the operation of positive selection (selection for beneficial mutations) in humans. This is precisely where Sanford's argument fails. In fact the opposite is the case: we have strong and abundant evidence that positive selection has occurred in the human evolutionary lineage. To Sanford's embarrassment , there has actually been a steady stream of papers demonstrating positive selection in the last few years, such as, Johnson et al. (2001), Sabeti et al. (2002), Nielsen et al. (2005), Chimpanzee Sequencing and Analysis Consortium (2005), and Voigt et al. (2006). To understand how these estimates work I would recommend evolgen's 7-part series of posts explaining how natural selection can be detected using molecular data (the last one is a good place to start). Only someone completely ignorant in the human population genetics literature could possibly claim that beneficial mutations don't exist in humans.

Since this is one of the central arguments in Sanford's book, I doubt that there is anything else worth discussing. However, I'll check the other chapters and will let you know if this is not the case.

Read on

What has Evo-Devo ever done for us? (II)

This discussion reminds me of a conversation I witnessed a few years ago between a population geneticist (PG) and an evolutionary developmental biologist (EDB) that shall remain nameless.

The EDB was explaining to the PG the significance of a paper that had made the cover of Nature a few years earlier. Those who are familiar with the literature on butterfly eyespots know that this is a classic of the field. It's beautifully written and illustrated. I know of more than one biologist that was turned on to the field of Evo-Devo by reading this paper. It shows how the expression of the regulatory gene Distal-less (better known for its role in limb development in Drosophila) is crucial to the formation of the eyespots of the wing of the African butterfly Bicyclus anynana. They analyse mutants, selection lines and different species to come up with a general hypothesis for how eyespot size and shape evolves in these butterflies. It has been cited 143 times and has sparked an entire research program in Evo-Devo.

The problem was that the PG didn't get what all the fuss was about. The EDB was getting more and more excited in trying to convey the beauty and elegance of the results, and he did so eloquently. At some point the PG said something like: "Of course I knew that some genes were involved. Is it really that important that we now know the identity of one of them?"

I think this exchange illustrates well the attitude of biologists for which the ultimate goal of evolutionary biology is not uncovering the precise steps involved in the evolution of a particular structure, but understanding the general evolutionary processes involved. Of course, evolutionary biology needs both.

Read on

Monday, October 09, 2006

Review of "The Mystery of the Genome" (I)

A few months ago, I had a debate with Salvador Cordova over robustness and redundancy. At some point (in these comments) Salvador decided that I needed enlightening on matters evolutionary and gave me a copy of a book by John Sanford, an agricultural geneticist at Cornell University (Courtesy Associate Professor), suggestively titled Genetic Entropy & The Mystery of the Genome (and ominously subtitled: "the genome is degenerating"). I suppose that courtesy dictates that I should say a few words about it. I finally picked it up and have read the first 30 pages (Prologue and Chapters I and II).

Unsurprisingly Sanford's objective is to demolish what he calls the "Primary Axiom" of evolutionary biology "that man is merely the product of random mutations plus natural selection". This is not a good start for someone who claims professional expertise in biology. Bringing the term axiom from mathematics into a discussion of a theory in the natural sciences is not helpful. What he describes is a summary of a theory, not an axiom. But let's go with it for the moment. So why might Sanford be challenging this "Primary Axiom"? He provides a telling answer:
If the Primary Axiom could be shown to be wrong it would profoundly affect our culture [...] It could change the very way we think about ourselves. (Prologue)
So now we know where he's coming from. What next?

In Chapter 1 ("The genome is the book of life. Where did it come from?") he explains how "the genome is an instruction manual". Although I would put it slightly differently, I don't have a major problem with his description: he talks of DNA, proteins, regulation of gene expression, etc. He then introduces a complicated metaphor -- manuals for constructing wagons -- for the process of mutation and natural selection designed to highlight the improbability of evolution. He goes into full "personal incredulity" mode:
Isn't it remarkable that the Primary Axiom of biological evolution essentially claims that typographical errors plus some selective copying can transform a wagon into a spaceship, in the absence of any intelligence, purpose, or design? Do you find this concept credible?
No prizes for guessing which answer he's counting on. While his wagon metaphor could have been more elegant (and more elegantly expressed) it is not fundamentally wrong. At least he doesn't tell us that natural selection is pure randomness. However, his discussion is misleading in at least two respects. First, it does not convey the immensity of time allowed for evolution to operate. This might be because Sanford is a Young Earth creationist and doesn't believe there has been much time for anything, although I haven't read a statement by him to that effect yet, so I'll give him the benefit of the doubt. Second, he analogizes mutation to spelling mistakes in natural language, which is deeply misleading. Although it is true that randomly altering letters in source code written in a high level programming language is unlikely to produce beneficial mutations, that does not imply that evolutionary computation is impossible, far from it. (Indeed, evolutionary computation raises major problems for evolution deniers, but that's another discussion.) The problem with Sanford's characterization is that point mutations are more subtle than the spelling analogy would suggest. But since no analogy is perfect, we'll let it pass for now.

Chapter 2 ("Are random mutations good?") gets to one of the central points of his argument. According to Sanford the bad news for the "Primary Axiom" is that "it can very reasonably argued that random mutations are never good". If true, this would indeed be a problem. So what about the evidence? Sanford tries to back it up with the following assertions:
  1. Mutations are like misspellings in the "instruction manual".
  2. There no "clear cases of information-creating mutations".
  3. The few beneficial mutations that occur are nearly neutral.
  4. Repeated selection experiments in plant breeding have resulted in "no meaningful crop improvement"
  5. Geneticists never see beneficial mutations.
Most of these are standard, repeatedly refuted creationist claims. I've already explained why the first argument is based on a misleading analogy. The second argument sounds more serious but it is hard to say what Sanford has in mind since he never precisely defines information, and creationists cannot be trusted to do so with any accuracy (see also this). In any event, mutations can create new binding sites, duplicate genes, etc, so whatever he means cannot be meaningful (to use one of Sanford's favorite terms). Let me give you an example from my study organism, the nematode Caenorhabditis elegans. A single aminoacid substitution in the gene npr-1 (neuropeptide Y receptor homologue) can make worms (normally solitary) aggregate on food. Is that an increase or decrease in information? Does it matter?

The fifth argument is completely wrong. Define beneficial and I'll give you many examples. Just in the nematode C. elegans we have mutations that increase or decrease body size, that increase or decrease lifespan, that increase or decrease hermaphrodite self-fertility, that make it easier or more difficult to go into dauer (the worm equivalent of a spore), etc. Many of these mutations can be beneficial in certain environments.

The fourth argument is nothing short of delusional. Artificial selection has succeeded in getting selection responses in the desired direction for "improvement" in practically every instance tried. For example, Ken Weber selected for differences in wing shape on the order of a few cells in the fruitfly Drosophila melanogaster and got a response! Exceptions to this generalization are so few and far between (e.g., changing the primary sex ratio and directional asymmetry in Drosophila) that the existence of constraints is still debated in the pages of Nature. There may have been a few unsuccessful selection experiments in crop species (I'm less familiar with that literature), but I doubt that Sanford's summary is accurate. Indeed I know of at least one case that contradicts it: starting with 163 ears of corn Leng (1962) was able to increase oil content of kernels from 4-6% to about 16% within 60 generations using artificial selection. That may not count as "meaningful crop improvement" in Sanford's book, but it does in mine.

We're left with the third argument. This one is more subtle and will be the subject of the next installment of this review. By then I may have been able to read a couple more chapters as well.

Read on

What has Evo-Devo ever done for us?

All right, but apart from the sanitation, the medicine, education, wine, public order, irrigation, roads, a fresh water system, and public health, what have the Romans ever done for us? (In Monty Python's Life of Brian, 1976)
PZ Myers has just posted a letter by Jason Hodin replying to a review of Sean Caroll's Endless Forms Most Beautiful that appeared in the New York Review of Books (which the NYRB refused to publish). Briefly, Hodin argues that Sean Carroll has hyped the field of Evo-Devo and that the NYRB reviewers (neither of which is trained in biology, apparently), went even further and practically attributed every advance in evolutionary biology from the past 150 years to Evo-Devo. I generally agree with both charges, but I am more interested in the former. While I respect Sean Carroll contributions to the field of Evo-Devo, I do not share his enthusiasm about the magnitude of evo-devo's accomplishments.

Jason Hodin argues :
I don't mean to denigrate my field of Evo Devo, nor do I intend to suggest that no critical insights have come from it. Perhaps the most important contribution of the field is methodological.
I agree with this sentiment. I believe that Evo-Devo has indeed crystalized a novel approach to evolutionary biology that is now changing research into other kinds of phenotypic evolution. However, we should not deceive ourselves into thinking that what Gould called "a new and general theory of evolution" is already here: it isn't. Understanding how certain developmental mechanisms have evolved adds more patterns to our picture of evolution, in much the same way as a new fossil does. Don't get me wrong: I'm in love with the details of the "new natural history" of Evo-Devo. But to trully revolutionize evolutionary biology, development has to be incorporated into the mechanisms of evolution, such as mutation and selection. Our current understanding of contraints, modularity and evolvability, for example, has not yet accomplished that.

Read on

Friday, October 06, 2006

That's Avida!

Mark Chu-Carroll has written an excellent post over at Good Math, Bad Math addressing creationist attacks on some research using the artificial life model system known as Avida. A descendant of Tom Ray's Tierra, Avida is used by several labs around the world, notably those of Chris Adami at Caltech and Richard Lenski at Michigan State University, to study a wide range of problems in evolutionary biology and ecology: from the evolution of robustness to adaptive radiation, from the evolution of sex to phylogenetic reconstruction. And, of course, the evolution of complexity.

On the face of it, work on these digital organisms (think of them as tame computer viruses) has grown into a vibrant research field in its own right, at the intersection between evolutionary biology and computer science. A quick search reveals at least 9 high profile papers on Avida in the last decade, including some of the ones I've already linked to. What do I mean by "high profile"? As a crude benchmark, those are papers with more than 10 citations each in other scientific papers (indeed, they were cited 37 times each on average). To put these numbers in perspective consider that the same database, the Institute for Scientific Information's (ISI) Web of Knowledge, tells us that William Dembski, none other than the Isaac Newton of information theory, has co-authored a total of 5 papers, cited 5 times in total (i.e., once each on average). So, there are Newtons and then there are Newtons...

So what do creationists think about all this work on Avida? Surely they would welcome an actual test of Behe's ideas on irreducible complexity, right? Wrong. They didn't like it one bit. So what did they do about it? Publish a rebuttal in the pages of Nature? Submit their own test to another journal? No. Instead they got Eric Anderson to sneer at the paper in one of their best journals, the Progress in Complexity, Information, and Design.

That might sound fair enough, but only if you don't know your scientific journals. The problem is that PCID, despite its impressive sounding description ("quarterly, cross-disciplinary, online journal that investigates complex systems apart from external programmatic constraints like materialism, naturalism, or reductionism") is not even up to the standard required for listing in the ISI. But is that really so bad? Well, the ISI currently lists 6088 science journals, including such obscure titles as Wool Technology and Sheep Breeding (impact factor 0.02), the Journal of the South African Institute of Mining and Metallurgy (impact factor 0.08) and Hazard Waste Consultant (impact factor 0, yes zero). So, yes, that is pretty bad. Publishing in journals not listed by the ISI counts for next to nothing in tenure decisions which could go some way towards explaining some of the problems Chapman listed in his pathetic apologia for the dismal state of scientific research on intelligent design.

And who, exactly, is Eric Anderson? The paper is not helpful in this respect, as no affiliation is listed. A search in ISI reveals no one of that name with a track record of publication in either evolution or computer science. For all we know, it might be a pseudonym for one (or more) of those "ID-friendly scientists" working at undisclosed locations that populate Chapman's dream world. Of course, anyone, credentialled or not, is entitled to comment on scientific matters. It's just that if you're going to fight evolutionary heavyweights like Lenski, Ofria and Adami, then you'd better make sure that you know what you're talking about. Specially when they have teamed up with Pennock, a philosopher, to make damn sure their argument was watertight. By now you've probably guessed it: Eric Anderson made a dog's breakfast of his critique. Let's turn to Mark for the details. Commenting on the abstract to Anderson's paper he says:
So the authors Avida are at best oblivious to the properties of the work they're doing; at worst, they're liars. And their work is based on on circular assumptions, which support Behe's notion of irreducible complexity. He's making an incredibly strong accusation against the Avida team: that either they're stupid and don't understand their own work; or they're liars.
In other words, not a promising start. And it gets worse. Not only does Anderson adopt an incredibly obnoxious and patronizing tone throughout, but he confuses hypotheses and assumptions and then accuses Lenski et al. of circular reasoning. Here's Mark's summary of the problem:

Now, here's where it gets really interesting. He says that he's going to examine the "key assumptions" built into Avida. But that's not really what he's going to do. What you'll see as I go through his paper is that he repeatedly tries to make it look like Avida is using circular reasoning. In fact, what they're doing is describing an experiment.

How do you do an experiment in real science? You start by developing a hypothesis. Using your hypothesis, you make a prediction. Then you perform the test, and see if the results match the prediction. If they do, then the experiment confirms the hypothesis (note, confirms not proves); if they don't, then the experiment disproves the hypothesis.

What the Avida team did was develop a hypothesis that an evolutionary system, working within the constraints of Behe's model of evolution could produce an irreducibly complex system. They proceed to describe their model, and the predictions it makes. Then they show their results, which confirm their hypothesis. Mr. Anderson tries to argue that because they stated their hypothesis up front, and then the test confirmed it, that they were cheating and being circular. He's pretending that the hypothesis is actually a set of assumptions; and that therefore, the experiment confirming the hypothesis is invalid.

Nicely put, hm? You get the idea.

Now, if you want to learn more about Avida, you could turn to Carl Zimmer's excellent piece in Discover. Here's one of my favorite passages:
One of the hallmarks of life is its ability to evolve around our best efforts to control it. Antibiotics, for example, were once considered a magic bullet that would eradicate infectious diseases. In just a few decades, bacteria have evolved an arsenal of defenses that make many antibiotics useless.

Ofria has been finding that digital organisms have a way of outwitting him as well. Not long ago, he decided to see what would happen if he stopped digital organisms from adapting. Whenever an organism mutated, he would run it through a special test to see whether the mutation was beneficial. If it was, he killed the organism off. "You'd think that would turn off any further adaptation," he says. Instead, the digital organisms kept evolving. They learned to process information in new ways and were able to replicate faster. It took a while for Ofria to realize that they had tricked him. They had evolved a way to tell when Ofria was testing them by looking at the numbers he fed them. As soon as they recognized they were being tested, they stopped processing numbers. "If it was a test environment, they said, 'Let's play dead,'" says Ofria. "There's this thing coming to kill them, and so they avoid it and go on with their lives."

This is yet another example of what has been called Leslie Orgel's Second Law: "Evolution is smarter than you are".

Read on

Monday, October 02, 2006

On Fire!

On a brighter note, C. elegans biology (my own real-life organism of choice) is on a roll. First, there was the Nobel Prize for the complete cell lineage in 2002. Now, the Nobel Prize for RNA interference. Did you know that it took John Sulston and his colleagues longer to complete the cell lineage work, than it took Fire and Mello to get the Nobel Prize since their paper came out. Congratulations!

Read on

Coming to a peer-reviewed journal near you... at some point... promise... trust us...

The Onion couldn't have spoofed the Discovery Institute better than its own President, Bruce Chapman. This statement is pathetic beyond belief.

[Via Red State Rabble]

Read on