[Stealth Hunter]

Transcription factors are and always will be an important part of molecular biology. By extension medical research for the foreseeable future. For the non-biologists among you, a transcription factor is basically a switch. It turns on a specific gene or set of genes in response to a specific stimulus. Hypoxia Inducible Factor, or HIF, as its name suggests, responds to a hypoxic environment and goes on to turn on, literally, hundreds of genes which induce changes in angiogeneisis, glycolysis, the Krebs Cycle, and oxidative phosphorylation, as well as reducing the metabolic load of the cell in question.

The transcription factor needs to know what gene to turn on, and where in the genome it's located. That's why in the 5' untranslated region of a gene which is to be activated (5'UTR) there is a specific site known as the transcription factor binding site, or consensus site. These sites are quite specific to the transcription factor in question, and induce transcription (and ultimately the translation) of the gene downstream of the consensus site.

If we know what genes are activated, how they're activated, and how specific transcription factors activate specific genes in a specific order, we can control the system and the potential for treatment of a host of diseases and injuries is enormous. For example, one of HIF's major activities is the growing of new blood vessels and capillary beds. In the heart, it remodels the myocardium and enhances cardiac function. If we can control what HIF does, we can potentially use the body's own mechanisms to repair the damage from a myocardial infarction with no ill-effects whatsoever.

Consensus sites tend to be short. Very short. On top of that, the binding factor doesn't bind to a unique site, but to a set of closely-related sites. HIF, to use my example above, has two known sites that it can bind with, both of which are 5 basepairs long.

Well, the problem with a short sequence is that it's not unique. A specific 25 basepair sequence is one that you'd only expect to show up once anywhere in the human genome. By comparison, if we have two possible sites, each of which has 5 basepairs, it shows up everywhere. Literally. To give you some perspective on what I mean by "everywhere," if you take our four bases, a, c, g and t, and construct a completely random strand of 2500 basepairs, along with its reverse complement, you'd expect one of our consensus sites to show up completely at random, at least once, on either the forward or reverse strands of that double-stranded DNA sequence. In the human genome, that means that you'd expect to see it over a million times just by pure chance.

The human genome only has about 30,000 genes, so obviously, all of those sites can't be real. But how can we look at a site and know that it's a real consensus site, or just one that shares its sequence?

Apparently, a couple of scientists, Loots and Ovcharenko developed a program called rVista in 2004. In the intervening four years, it has become the way in which one differentiates a "real" consensus site from a "fake" one.

You start with the assumption that a number of different animals share a common ancestor. You look at the gene sequence of animals closely related to it on the evolutionary tree. If you're studying a mouse, you compare your sequence to that found in rats and rabbits. If you're studying a human, you compare your sequence to Chimpanzees and Orangoutans. The program uses two criteria to filter the "fake" sequences from the "real" ones. 1) functional sequences will be conserved by evolution. 2) functional sequences will be more accurately conserved between closely related species than between distantly-related ones. From this, it aligns the sequences and determines whether these short, non-coding consensus sites are significant.

It works. It has a lower than 4% false-positive rate. It would only work if evolution were true, because the program is designed to analyze sequences under an evolutionary framework. If evolution is false, then the program would be unable to produce any valid data whatsoever.

The process by which rVista was developed and desiged can be found in:

G. Loots and I. Ovcharenko, rVista 2.0: evolutionary analysis of transcription factor binding sites. Nucleic Acids Research, 32, W217-W221 (2004)

It's somewhat technical, but as you can see, this algorithm is completely useless, unless evolution in general (common descent to be specific) is true.

Now my question: What would be the Creationist approach to the same problem?