This post continues directly from Part 1.
Koonin is clearly very impressed with the new paper, which he calls “brilliant” and “a genuinely important work that introduces a new and potentially major mechanism of evolution…” His enthusiasm is a good indication that the problem is a major one, and that no other papers exist which deal effectively with it.
So what is the paper (a theoretical, mathematical-modeling study) about? When a mutationless gene is transcribed and translated into a protein, errors can creep in. It turns out that these error rates are much higher than for copying DNA. Using published mutation rates, Whitehead et al (2008) estimate that 1 in 10 standard-sized proteins will contain an error; that is, they will contain an amino acid that is not coded for by the gene. The authors call these “phenotypic mutations.” Inherited changes that occur in the DNA are called “genotypic mutations.” Now, the idea is this. Suppose an organism needs two mutations to acquire some new feature, such as a disulfide bond. Further suppose that a single organism in a population that initially has neither of the mutations acquires just one of the necessary mutations in its DNA. Because of phenotypic mutations, this single organism will also contain some copies of the protein that have the second mutation. If the selective benefit of these phenotypic mutations is proportional to their concentration, as the authors suppose, then that organism may have an advantage over other organisms with no mutations. In a sense, the authors say, evolution can look a step ahead, so the authors dub this the “look-ahead effect.” The reviewer Eugene Koonin agrees that the paper “in a sense, overturns the old adage of evolution having no foresight. It seems like, even if non-specifically and unwittingly, some foresight might be involved.”
As the authors and one of the other referees note, this is pretty reminiscent of something called the “Baldwin effect”, which was first proposed in the 19th century. The authors contend that there are subtle differences between the Baldwin effect and the look-ahead effect. Yet, whoever deserves priority for the idea, I don’t think the look-ahead effect contributes much at all to solving the problem of multiple mutations. In my own opinion, the idea of the paper is certainly clever, but Koonin vastly overestimates its importance. It offers virtually no help in solving the “old enigma,” as I explain below and in Part 3.
First, the effect is quite minor at best. Since, based on transcriptional and translational mutation rates, the fraction of proteins with the correct phenotypic mutation is expected to be about one-hundredth of one percent (10^-4) of the total number of protein copies, the presumed selective effect will be only 10^-4 times the selective effect of the double genotypic mutant. So if the double genotypic mutant had a selective advantage of 0.1 (a pretty substantial value), the phenotypic look-ahead mutant would have an advantage of just 10^-5. If the double genotypic mutant has less of an advantage, the look-ahead has proportionately less. Because of this, the effect would be helpful only for large population sizes: too small of a population and there is no effect, because the mutation is effectively neutral. One can construct situations in which the selective advantage of a particular double genotypic mutant would be enormous (for example, if it conferred antibiotic resistance) so the look-ahead effect would be greater, but positing the general occurrence of such situations in nature amounts to special pleading.
It’s also important to realize that the authors of the paper purposely did not consider mitigating factors in their analysis. As they wrote, “The goal of our analysis was to demonstrate that the look-ahead effect is theoretically possible, and as such, we intentionally excluded confounding factors for the sake of clarity.” Other possible important effects that weren’t considered in the model include the influence of the first genotypic mutation on the stability of the spectrum of proteins with phenotypic mutations, effects of the mutations on translation rates, and so on. It is certainly understandable to simplify a model as much as possible for an initial investigation. However, any confounding effects will only diminish the strength of an already-weak influence.