In his last comment on the question originally raised by Bill Dembski (“Is James Shapiro a Design Theorist?“), Jim Shapiro writes:

As my book details, we have many examples of this process [domain shuffling and accretion] mediated by mobile genetic elements in nature as well as its replication in living cells in the laboratory.

Ann Gauger and I are familiar with Shapiro’s thinking on this subject, and we have a high regard for his contributions.
We think much of the disagreement on protein origins stems from the fact that two very different approaches are being taken, both of which Shapiro refers to in the sentence quoted above. The first approach, which dominates the scientific literature, is to make inferences about how things evolved by looking for clues in genomic sequences. Hypotheses are formulated on the basis of these clues, and then “tested” by combing through the sequence databases in search of more examples that fit the hypothesis.

The second approach is, as Shapiro says, demonstrating that a hypothesized mechanism actually works in living cells.
Clearly, these are not equivalent exercises. The first is a valid way to formulate hypotheses about protein origins, but the only convincing way to show that a hypothesized mechanism really works is to observe it working in the present instead of merely inferring that it worked in the past.

As people who have specialized in this second approach, Ann and I are of the opinion that it is widely neglected in the evolutionary literature. We are well aware of the many experimental studies that attempt to draw some conclusion about protein evolution, but we also know that very few of these studies take the critical approach of asking whether the presumed evolutionary mechanism really works. It’s as though most scientists are unwilling to put these fundamental ideas to a serious test.

I can make this point most clearly with a concrete example. We can go into the lab and modify bacterial cells by deleting the entire set of genes dedicated to the synthesis of tryptophan, one of the essential building blocks of proteins. When we observe what happens when these modified cells are given just enough tryptophan to grow and reproduce, we will see lots of things happening, but none that can be expected to reinvent a set of genes for making tryptophan, even in a large population over billions of years.

I know of many processes that people talk about as though they can do the job of inventing new proteins (and of many papers that have resulted from such talk), but when these ideas are pushed to the point of demonstration, they all seem to retreat into the realm of the theoretical. Having followed this debate for some time now, and having made several experimental contributions to it, Ann and I have become convinced that none of the current naturalistic ideas about the origin of protein folds or the functional diversification of existing folds actually works in any general sense.
But of course, as experimentalists we are very willing to see the evidence that might prove us wrong.