11 years ago (give or take), I posted an essay on The Panda’s Thumb that discussed what I believed to me some serious flaws in Doug Axe’s 2004 paper in the Journal of Molecular Biology, or rather, flaws in the ways that the ID community have used the conclusions from that study. Over the years, there have supposedly been some responses from Axe to my essay, but in all reality, to my knowledge, Axe has never addressed the central criticisms I discussed. Joshua has been kind enough to give me a place to revisit this, and maybe elicit a relevant response from the ID community – perhaps Axe, perhaps Anne Gauger, perhaps Brian Miller (the latter two of whom have both avoided my essay with what amount to ad hominem remarks.) What I will do here is: to as briefly as possible, recapitulate my central criticisms so that readers here may understand some of the issues; and raise some new concerns that arose (in my mind) a few years ago, after reading another of Axe’s papers.
For starters, let’s recall the task at hand – to estimate the fraction of all sequences that may possess some sort of enzymatic activity. In my essay, I laid this problem out with a picture – Fig. 1 in the essay:
Basically, the height of the hill I draw represents some quantitative measure of enzyme activity, and the base of the hill represents the total number of sequences that may possess at least a modicum of activity. In other words, the fractional area of the base of the hill compared with the entirely of sequence space can be thought of as a measure of the fraction of sequences that have function.
My criticism, simply stated, is this: In his paper, Axe did not study the “hill” shown in Fig. 1; my reasoning for this statement is spelled out in his own words in the JMB paper and, I believe, accurately recapitulated in my essay. Rather, he studied an enzyme with very low activity and deliberately designed (in the real sense of the word) to be very sensitive to mutation. This variant is depicted (very approximately) in Fig. 3 of my essay:
All this is well and good, and it is accurate to state that Axe’s experiments may* have estimated the base of the hill shown in Fig. 3. However, in all of the ID literature that refers to this study, including work by Axe himself, it is assumed that the base of the hill in Fig. 3 is the same as that in Fig. 1, and that Axe’s work actually directly measures the base of the hill in Fig. 1. This is patently false. In order to generalize even a little bit (say, to enumerate all sequence variants that may possess the activity Axe is supposedly measuring), some sort of conversion factor is needed. None has ever been offered by anyone in the ID camp, as far as I know. Moreover, no ID proponent – not Axe, not Gauger, not Miller, anyone else – has ever addressed this criticism. The responses I am aware of range from simple avoidance to blatant ad hominem. What I hope to accomplish here is to wrap up this debate, to get some sort of serious and relevant response that directly and properly addresses my criticisms. In so doing, I am glad to explain my reasoning in whatever detail participants here care to put up with.
*As I said above, in reading some of Axe’s later work, I realized that the assays he used in his JMB paper are not really suited to the task at hand. In all of the work I am aware of, Axe uses growth of E. coli on ampicillin as a proxy for beta-lactamase activity; as far as I know, he has never published results of direct enzyme assays of any of the beta-lactamase variants he is studying. (I am more than happy to be corrected in this regard, and would be glad to update the following with the data I am asking for below.). Anyone who has plated transformed E. coli on ampicillin or other beta-lactams knows that this isn’t a very quantitative method, and certainly not suited to quantitatively distinguishing enzyme variants with differing activities. More to the point of the JMB paper, Axe’s BioComplexity paper shows that, in cases where enzyme activity may be very low, this assay cannot reliably distinguish between low and zero enzyme activity. This is very important – I believe the conditions used by Axe in the 2004 JMB paper cannot distinguish between mutants with low but significant activities and those with no activity. In other words, the accuracy of Axe’s estimates in the 2004 paper cannot be known, and he may be wrong by tens and tens of orders of magnitude or more. (To illustrate this, readers can replace the fraction of active variants that Axe uses in his calculations with, say, 0.9. Since Gauger and Miller don’t think I am competent enough to do this, readers here are invited to do this and report back on their findings. There are lots of smart physicists and engineers here who should be up to this task.). At the very least, Axe needs to directly measure the enzymatic activities of his “reference” and tens (at the least) of active and presumably inactive mutants isolated in his study. Not by plating E. coli on ampicillin, but by performing proper, rigorous, and quantitative measurements of beta-lactamase activity.
These are, I believe, pertinent and serious criticisms. I am more than happy to explain this essay and my original one for participants here, and am hopeful that Miller and Gauger (at the very least, since they have offered quite rude and irrelevant criticisms of this essay in the past) and perhaps Axe might join us to hash out these issues.
Postscript - Axe’s 2004 paper remains, as far as I know, behind an onerous paywall. I am glad to share it privately with participants here, if you are comfortable with contacting me at my uni email address. I won’t post the paper or email here, but I am pretty easy to track down.