I’d seen it before but never read it through. In response to your question here, though, I decided to read it in detail (skimming some of the boilerplate about frequentist and Bayesian statistics and the like). Holy cow – it’s worse than I thought. Pretty much all of the work that they cite to support their conclusions comes from a handful of ID and creationists sources, while many of the references to the conventional scientific literature seem to be window dressing.
Take section 4.1. There, the authors center their evaluation of the rarity of functional proteins on a single study by Axe (a Discovery Institute guy) on beta-lactamase, which is one source of antibiotic resistance. That study concluded that much less than 1 protein in 10^50 would have beta-lactamase activity. Axe’s approach has grave problems and even in principle is simply not capable of estimating what they’re citing it to estimate. More importantly, it’s flatly contradicted by other experimental work, e.g. a study of a random peptide library that found three with beta-lactamase activity out of ~10^9. That’s a difference of 41 orders of magnitude.
Thorvaldson and Hössjer do cite (even if they actually cite the wrong paper) a much better study that searched for a different specific function (ATP binding) among random proteins. This study found that 1 protein in 10^11 had the desired function, again vastly higher than the estimate accepted these authors end up using. The lower estimate is consistent with other studies, like the beta-lactamase one I mentioned above, that have found functional proteins occurring in the ballpark of once per 10^10, give or take a couple of orders of magnitude. Even though they cite this study, though, the authors don’t seriously engage with it or with any of the rest of the extensive literature on the subject. Instead, they immediately dismiss it all, citing one study – which turns out to be a preprint (never published anywhere) written by two creationists and posted on a preprint server devoted to fringe science. This is not how one does science.
This is not some minor point: how rare functional proteins are is central to the ID claim that new proteins should only very, very rarely be able to evolve naturally. Axe’s estimate would, if accurate, really be problematic for evolution, while the estimates coming from mainstream science aren’t. (Note that these estimates are for randomly finding one or a few specific functions; given the vast number of possible functions, finding any functional protein is much easier.) Oh, another interesting study. One researcher tried looking for specific functions by randomly combining fragments of existing proteins, rather than creating them from random sequence, and found them at a rate of more than 1 per 10^5.
It’s a good thing for all of us that functional proteins are not really that hard to find: we’d all be dead if they were. Our adaptive immune system deals with pathogens its never seen before by developing antibodies, which are proteins with the specific function of binding to the invader. Out of a repertoire of something like 10^8 existing antibodies, which were generated by a process of random mutation, our bodies consistently find a functional match to each invader, which they then refine by means of further mutation and selection. The result is an exquisitely tuned protein, produced by random processes in less than a week with no intelligent intervention.
What I mean about window dressing can be illustrated by this passage, which is about the probability of configuring multiple proteins to work together: ‘P(Ac) can still pose huge obstacles to the chance configuration of the quaternary structure of operative biological systems (Csermely et al., 2010). This problem of estimating P(Ac) seems quite intractable, but it may be addressed by performing perturbation experiments (Antal et al., 2009). The idea is to take a functional system, perturb it, and determine how perturbation affects the probability of retaining function.’ Nothing in the Csermely et al. study says anything about there being huge obstacles to chance configuration of the parts, and the summary of Antal et al. is quite wrong: the perturbation approach in that paper has nothing to do with evolving proteins or the probability of retaining function – it’s about modeling the dynamic behavior of complex protein networks in a living cell. These citations really have nothing to do with the statements they’re supposed to be supporting.
I quit reading when I got to this: ‘Haldane (1932) asked the question whether natural selection is capable of removing deleterious mutations as they arrive within a species over time. If not, they may cause a mutational load that increases to such an extent that the survival of the species is threatened (Lynch et al., 1993). Such an increased mutational load corresponds to an increase of genetic entropy (Sanford, 2008)…’ Just about everything here is wrong. That’s not what Haldane was asking, Lynch’s paper only applies to asexual species with tiny populations, and Sanford’s genetic entropy is about mutations whose effect is too small to be removed by natural selection, which is a different issue. Genetic entropy also happens to be a crackpot idea that has zero evidence to support it and abundant evidence refuting it, and that only makes sense in a young earth creationist context.