Dembski Responds to Rosenhouse

[Rumraket (Mikkel R.)]

@lee_merrill

Here is the crux of the argument:

“Let’s make a rough calculation for the average number of organisms we would have to slog through to find a new protein-protein binding site. As I said, shape space tells us that about one in ten to a hundred million coherent protein-binding sites must be sifted before finding one that binds specifically and firmly to a given target. The simplest way to alter a protein is by point mutation, where one amino acid is substituted for another at a position in a protein. There are twenty different kinds of amino acids found in proteins. That means that if just five or six positions changed to the right residues—the ones that would allow the two proteins to bind—that would be an event of approximately the right frequency, since twenty multiplied by itself five or six times (20^5 or 20^6) is about three million or sixty million, respectively—relatively close to the ten to a hundred million different sites we need. So one way to get a new binding site would be to change just five or six amino acids in a coherent patch in the right way. [12] This very rough estimation fits nicely with studies that have been done on protein structure. [13]” (Edge of Evolution, pp. 133-134)

Thanks. Unfortunately there are many things either wrong or unclear with what Behe says there.

@lee_merrill

Behe: Let’s make a rough calculation for the average number of organisms we would have to slog through to find a new protein-protein binding site. As I said, shape space tells us that about one in ten to a hundred million coherent[as opposed to incoherent?] protein-binding sites must be sifted before finding one that binds specifically and firmly to a given target.

In the first sentence Behe says he’s trying to calculate how many organisms we have to “slog through” to find a new protein-protein binding site. It’s not clear what the hell that means. Are we supposed to just draw organisms at random and see if any one of them has a protein-protein binding site not found in one of the other organisms? Already at the first sentence there’s problems. What he says is unclear, it’s unclear why he says it, and it’s inconsistent with what he says later when he starts talking about sifting protein-protein binding sites.

Then in the next sentence he asserts that “shape space” tells us something, but without a reference. He says he’s said it before, but you don’t quote him doing that, so now we have to go look at what he presumably(?) said previously that show how “shape space” tells us this thing he claims it does. Then there’s the “coherent” thing. What’s a coherent protein-protein binding site as opposed to one that’s… not coherent? I don’t understand what the word coherent is doing in that sentence, it’s not used before and not after in what you quote.

Then there’s the problem with target thinking. Now suddenly we’re not just looking for a new protein-protein binding site—which he said to begin with—now we’re looking for a particular target to be bound. No. Evolution doesn’t have targets. When new functions evolve, whether they’re protein-protein binding sites or something else, it’s because those were the “solutions” that happened to get sampled by mutations(of all the types that occur) and recombination, that also happened to improve fitness.
It doesn’t make sense to pick a specific result out and compute it’s odds. For the same reasons I explained in the T-URF13 thread, target thinking is extremely misleading.

So to determine whether it is difficult to evolve protein-protein binding sites in general, the question is what is the probability of any binding site evolving, rather than a specific one. And then whether selection can aid in that process should such binding be advantageous(which we can’t really know a priori). I’ve already cited literature that shows it is easy to select for improved binding, meaning when and if the association between two proteins is advantageous they quickly turn into strong binding spots.

@lee_merrill

Behe: The simplest way to alter a protein is by point mutation, where one amino acid is substituted for another at a position in a protein. There are twenty different kinds of amino acids found in proteins. That means that if just five or six positions changed to the right residues—the ones that would allow the two proteins to bind—that would be an event of approximately the right frequency, since twenty multiplied by itself five or six times (20^5 or 20^6) is about three million or sixty million, respectively—relatively close to the ten to a hundred million different sites we need.

Bzzzt. Stop. Why does Behe think we are five to six mutations away from a binding site, instead of one or two? Why does he appear to think only one residue out of the 19 alternatives at each of those five to six positions produces effective binding?
That is extremely dichotomous thinking and he should know better as a person with a goddamn PhD in biochemistry. Binding is not all or nothing, different residues show different degrees of affinity to each other. His assumption here seems to be just invented out of thin air to make it seem as difficult as possible to evolve novel protein-protein binding spots.

We know from concrete empirical evidence that merely increasing hydrophobicity by as little as a single mutation can trigger strong and effective binding in many locations on a protein’s surface. I have already cited you literature that details experiments that show this.

@lee_merrill

Behe: So one way to get a new binding site would be to change just five or six amino acids in a coherent patch in the right way. [12] This very rough estimation fits nicely with studies that have been done on protein structure. [13]

Now with a coherent patch. What’s that? What’s an incoherent patch? Does he mean some area on the surface of a protein? How are areas coherent or incoherent? And I’d very much like to see references 12 and 13. Rule of thumb: When ID proponents cite literature then check the references.