Intelligent Design creationism is bad theology, bad politics, bad education, and bad science. That last point is made for me every day as I read the real science literature, and see what a contrast it makes with the ideological press releases that come out of the Discovery Institute. In particular, as I was reading a recent review article by Peel (2004), I was struck by the way scientific work builds on past observations, integrates multiple lines of evidence, and makes justified predictions about the natural world that are amenable to testing…all things deplorably absent from ID creationism.

The common theme in ID creationist research seems to be an assertion of the negative: science can't explain X. Y is an impenetrable barrier. Z can't possibly happen. You can't get here from there. Dembski, Behe, Meyer, and Nelson are all taking this approach, and worst of all, justifying it by carefully omitting all the evidence that shows that X can be explained, Y can be crossed, Z did happen, and of course we got here from there. It's also a failure as a research program, because their point of view is utterly dependent on not finding evidence.

So let's take a look at how scientific minds deal with an awkward problem in evolution.

Here's the situation: the paradigmatic example of early pattern formation in the development of the embryo comes from Drosophila. We know lots about the molecules and interactions in the fly embryo that set up a head and tail end, and partition the animal into segments. We know far more about these processes in the fruit fly than any other organism on the planet—there are gradients of molecules that induce localized regions of gene expression (by the gap genes) that then lead to the expression of alternating stripes of genes, the pair-rule genes, and then to the expression of genes in subregions of every segment. It's beautiful stuff. Here's a diagram of the hierarchy of expression of these genes in the fly, from gap genes to pair-rule genes that are turned on in every other segment to segment polarity genes that are active in domains within every segment.

Head versus trunk segmentation in Drosophila. Head and trunk segments are patterned by different segmentation gene cascades during Drosophila development—the two cascades involve different gap genes, pair-rule expression is a unique feature of trunk segmentation and segment-polarity gene expression appears later in the head. However, the discovery of collier suggests that there is at least one gene acting at a level similar to the trunk pair-rule genes within the head segmentation cascade. The two cascades overlap and interact within the mandibular segment. Curiously, the mandibular engrailed stripe is still expressed when the first stripes of the pair-rule genes hairy and even-skipped are missing. Instead, buttonhead is required for engrailed expression, although it is unclear whether this regulation is direct. buttonhead is known to regulate even-skipped during cephalic furrow formation, and even-skipped stripe 1 positions the posterior border of collier expression. The red dashed line marks the parasegment boundary that is homologous to the border between naupliar and post-naupliar regions in the crustacean Orchestia cavimana.

There is a problem in this pretty picture. The answers we know in flies can only apply to flies. Flies have several oddities, not shared by many other animals, that are necessary for the particular molecular functions in this pattern of development to work. For one, flies are long germ band insects; that means that all of their segments develop at roughly the same time, so the stripes of their segments crystallize out relatively suddenly. Other insects have short germ bands, and add their segments sequentially, from anterior to posterior. We vertebrates also add our segments one at a time. Another important difference is that the fly embryo is initially syncytial—nuclei proliferate, but no membranes form between cells. That means there are no membranous barriers between cells, and pattern-forming molecules can freely diffuse between them. Most other animals are cellular at these same stages. That means we require extracellular molecules and complex signal transduction mechanisms to transmit information from one cell to another.

The creationist answer to all this is predictable: you can't get here from there, and we didn't evolve from a common segmented ancestor. Therefore, evolution is false, and we don't expect any necessary historical linkage between these different lineages. We can stop looking.

The scientific answer is rather different. We need to explore other systems than flies, and dig a little deeper to see what their similarities might be. We should make testable predictions about possible linkages and do experiments to see how well they hold up. Interesting and complex differences are, well, interesting—they drive further research into the problem, not surrender.

In Peel's paper, he reviews what we know about segmentation in various organisms, and uses those observations to make a series of predictions about how the highly derived segmentation rules of Drosophila evolved from a simpler pattern, still extant in other insects and in us vertebrates.

I...present a model for the evolution of the Drosophila paradigm. The model predicts that the primary pair-rule genes of Drosophila ancestrally functioned within and/ or downstream of a Notch-dependent segmentation clock, their striped expression gradually coming under the control of gap genes as the number of segments patterned simultaneously in the anterior increased and the number patterned sequentially via a segmentation clock mechanism in the posterior correspondingly decreased.

The pattern that we see in vertebrates, as a counterexample to what goes on in flies, is that most segments form sequentially, from front to back, and that it all occurs when the tissues are fully cellular. Cells use a clock. A set of genes in the Hairy/Enhancer of split related (HER) family undergo rhythmic oscillations in their expression, which, together with a receding wavefront of FGF8 and Wnt signals, is used to set aside clumps of cells that will form the segment. Another set of well-known genes, Notch and Delta, are coupled to this cycle, and are believed to be important in synchronizing adjacent cells. Keep those names—Hairy, Notch, and Delta—in mind. They come up a lot when we're looking at segmentation in different organisms.

There are a couple of pieces to the story.

One is that we see differences in the segmentation rules within the fly. The head seems to segment differently than the trunk and abdomen, and although relatively poorly studied, the regulation of head segmentation so far seems to be more highly conserved in the arthropods.

Another is that pair-rule gene is probably primitive. Pair-rule stripes of expression pop up in many lineages. What can't be the same between flies and other organisms, however, is the mechanism that sets up the striped pattern, for reasons listed above.

The cis regulatory elements of the pair-rule genes in flies are extremely complex, containing elements necessary to position each stripe of expression along the length of the animal. The regulatory elements of the pair-rule genes in animals that use a clock are simpler.

We see the same molecules in all of these lineages, too. Hairy is a pair-rule gene in flies (click on the color diagram up above to see a bigger image: there's Hairy in the fly!) Hairy has also been found in spiders and vertebrates; other fly pair-rule genes, such as runt, paired, and even-skipped have also been found in spiders. Notch and Delta are more or less ubiquitous.

Putting all this together, what Peel hypothesizes is that the pattern we see in vertebrate segmentation is much closer to the ancestral condition than what is going on in flies. In Drosophila we see the end result of millions of years of selection for more rapid development. As the time allowed for segmentation was compressed, the clocklike model was progressively abandoned from front to back. The gap genes were adopted segment by segment to regulate the pair-rule genes, until the clock was rendered completely superfluous.

This is more than a just-so story. It's built on comparative data, such as knowledge of the genes involved in segmentation in different lineages, and the existence of invertebrate intermediates that segment different degrees of their body plan sequentially. Most importantly, it provides a set of specific, testable predictions:

• Short-germband arthropods will use a segmentation clock, with rhythmic oscillations of pair-rule and Notch-related genes.
• There will be a gradient of a signalling molecule (most likely an FGF/Wnt) on the longitudinal axis of short-germband arthropods.
• Segmentation in organisms that add segments sequentially will not be dependent on any of the gap genes, which are a derived function added in Drosophila evolution.
• The regulatory regions of pair-rule genes in short-germband insects will lack the stripe enhancers we see in flies, and will instead have clock enhancers that are sensitive to Notch signalling.
• To account for known differences between vertebrates and invertebrates, the arthropod clock should tick off two segments for every oscillation, where the vertebrate clock ticks off just one (I won't go into the logic for this, but the diagram below explains what Peel expects).

Vertebrate clock versus a putative arthropod segmentation clock. A: Each oscillation of Hairy/enhancer of split related gene expression in vertebrates gives rise to one somite. The horizontal bars represent paraxial mesoderm separated in time by one full cycle of HER gene oscillation. B: If Hairy once functioned as an oscillator within an arthropod segmentation clock mechanism, the evolution of its pair-rule expression in Drosophila can most easily be explained if one oscillation of the ancestral clock patterned two segments worth of tissue. The horizontal bars represent an arthropod germband separated in time by half a cycle of Hairy oscillation. HER/Hairy expressing tissue is shaded. Note that Hairy pair-rule stripes of expression in Drosophila do not correspond exactly to future segmental borders, (13) and so may not have done so in ancestral arthropods, as implied by the figure.

You see, this is the kind of thing scientists are looking for in their research models: ideas that tie together many observations in a coherent, specific way (no, "a designer did it" is not very specific), and that provide a foundation to guide more work. Intelligent Design fails on both counts. Peel's theory could turn out to be wrong (I wouldn't bet on that) or overly simplified (that wouldn't surprise me at all), but it's still a genuine scientific framework, and we'll learn something by evaluating it.

Peel A (2004) The evolution of arthropod segmentation mechanisms. BioEssays 26:1108-1116.