Best thing is Behe & Snoke 2004 basically admit that their paper commits the Texas Sharpshooter fallacy in the discussion(my bolds):
Such numbers seem prohibitive. However, we must be cautious in interpreting the calculations. On the one hand, as discussed previously, these values can actually be considered underestimates because they neglect the time it would take a duplicated gene initially to spread in a population. On the other hand, because the simulation looks for the production of a particular MR feature in a particular gene, the values will be overestimates of the time necessary to produce some MR feature in some duplicated gene. In other words, the simulation takes a prospective stance, asking for a certain feature to be produced, but we look at modern proteins retrospectively. Although we see a particular disulfide bond or binding site in a particular protein, there may have been several sites in the protein that could have evolved into disulfide bonds or binding sites, or other proteins may have fulfilled the same role. For example, Matthews’ group engineered several nonnative disulfide bonds into lysozyme that permit function (Matsumura et al. 1989). We see the modern product but not the historical possibilities.
So Behe & Snoke model a situation in which only a particular set of mutations can yield the function of interest, that no other similarly complex adaptations are possible, requires 2 or more specific mutations (in other words that only 1 out of 20 amino acids are allowed at specific sites), and in which intermediate single mutations are deleterious. And that the novel set of specific mutations must evolve in a gene not under purifying selection before a null mutation occurs. And they assume the majority of mutations cause a null function that renders any subsequent mutations that might have been part of the novel function to be useless because the gene is now broken.
How Behe can admit this much in the very paper, and still go on to pretend his model constitutes some sort of problem for evolution is astounding.
It gets worse. Other absurd features of the model:
In this paper, we report the results of the stochastic simulation of the time to fixation of new MR features by what we consider to be the conceptually simplest route: point mutation in the absence of recombination in a duplicated gene that is free of purifying selection.
And:
The model presented here assumes that newly duplicated genes encode a full-length protein with the signals necessary for its proper expression. It is further assumed that all duplicate genes are selectively neutral.
And:
The pertinent feature of the model is that multiple changes are required in the gene before the new, selectable feature appears. Changes in these nucleotide positions are assumed to be individually disruptive of the original function of the protein but are assumed either to enhance the original function or to confer a new function once all are in the compatible state. Thus, the mutations would be strongly selected against in an unduplicated gene, because its function would be disrupted and no duplicate would be available to back up the function.
These two assumptions, that the individual mutations are deleterious in the functional copy, and that purifying selection is not operating on one of the copies, combine in an extremely restrictive way. Because one of the copies of the gene is assumed to be free of purifying selection, mutations that degrade the original function are not selected away. And these are of course assumed to vastly outnumber the productive mutations that can lead to the new function, so the probability is one of the two copies will sufer a null mutation way before one of the rare novel function-contributing mutations occur.
It’s absurd.
The basic “task” that the model asks a duplicate gene to perform is to accumulate mutations at the correct nucleotide positions to code for a new selectable feature before suffering a null mutation.
And:
However, if several point mutations (indicated by a “+” in the figure) accumulate at specific nucleotide positions (indicated by the three squares outlined in blue in the
figure) in the gene coding for the protein before a null mutation occurs elsewhere in the gene (indicated by a red “X”), then several amino acid residues will have been altered and the new selectable MR feature will have been successfully built in the protein (indicated by the greenshaded area). By hypothesis, the gene is not selectable for the new feature when an intermediate number of mutations has occurred, but only when all sites are in the correct state.
This means Behe & Snoke are modeling a situation where they are demanding a particular set of mutations to evolve to create a new function, with no alternative mutations being able to create a similarly complex function, and that this occurs in one of two copies of a gene where vastly more likely deleterious mutations that render any subsequent mutations invalid, are not purged by selection.
Bill thinks this is a more realistic model of protein evolution.