Review of William B. Provine’s “The ‘Random Genetic Drift’ Fallacy”

Just about a year ago, I got an unusual email. It was from William Ball Provine, a well-known historian of science most famous for his work on the modern evolutionary synthesis. Provine was especially well-known to me as the Ph.D. advisor to Greg Graffin of the punk band Bad Religion; although this is a story for another time, one could make the argument that my own intellectual interests are phylogenetically-related to Provine via Graffin, as it was evolution-infused Bad Religion lyrics coming through a Discman during my early-1990’s days as an undergraduate biology major that first got me excited about evolution. Why was this really prominent historian of science contacting me, a novice science blogger who is lucky if fifty out of earth’s seven billion people visit his site each day? Provine’s humble email was pretty simple:

Although I knew very little about the book, not surprisingly I was intrigued. I agreed to read the book and review it on my site, but confessed that I would not be able to start reading it for several months. Dr. Provine sent me The “Random Genetic Drift” Fallacy several weeks later: a simple, short, self-published book with a cover that reminded me slightly of a Princeton Monograph. I started reading the book in April and finished it off in late June. I feel bad for taking so long to write it a proper review, but this sort of delay (or worse) is unfortunately typical for me.

Sadly, I learned yesterday morning — just as I sat down to write this review — that Dr. Provine died in early September of this year after two decades of living with a brain tumor [1, 2, 3]. His book has provoked so many thoughts, and I am saddened to learn that I can no longer send my questions his way. Rather than being a polished manuscript, The “Random Genetic Drift” Fallacy reads like a treatise written by a man with something very serious to get off of his chest. But what ideas Provine had to get off of his chest! Although I know none of the details of Dr. Provine’s health problems, the fact that he was suffering from a terminal illness provides a little context: these are the last formally-published ideas of a man who probably knew that he was soon to die.

It takes some serious conviction to be willing to put an idea in quotes throughout an entire 178-page book just because you want to make it abundantly clear to your reader that this idea is fallacious. That’s the kind of conviction Provine expresses in this book: “random genetic drift” is an idea that needs to be purged from our thinking about how evolution works, and if that means putting the term in quotes every time you use it, so be it. The book vigorously questions assumptions made at the very foundation of population genetics and in doing so suggests that the oft-lauded “modern synthesis” — theoretical work that explained the genetic basis of evolutionary change — might need to be reconsidered. If what Provine has to say is apt, perhaps the job of synthesizing Darwinian evolution and genetics is not done: we need a “moderner synthesis”. Is this an important book with the potential to revolutionize our thinking about evolutionary genetics or a misguided and behind-the-times critique of past history rather than current science?

Everything you know about population genetics is wrong?

Provine’s central thesis is that the idea of random genetic drift emerges from erroneous assumptions. Worse yet, because the phenomenon of genetic drift creates a null hypothesis for population genetics, the entire foundation of the field is unstable. Provine knows the early history of population genetics well, and provides an intriguing narrative on the contributions of R.A. Fisher, J.B.S. Haldane, and Sewall Wright. Provine was especially knowledgable about Wright, as he pored over Wright’s notes and personal papers in writing his 1986 book Sewall Wright and Evolutionary Biology. Starting with Fisher, Provine provides a compelling argument that during this influential period, the realities of biology were subjugated to the convenience of theory. In particular, the founding fathers of population genetics wanted to import the statistical modeling approaches that had proved so successful in physics into the science of genetics. Doing so required making a crucial assumption about alleles at a particular gene locus: each was treated as an independently-evolving entity despite being attached to other alleles at other gene loci on the same chromosome.

Anyone who has studied evolution has seen the marble metaphor, which is a physical model of the way that genes are treated in the population genetics models of the modern synthesis:

Image courtesy of Wikimedia Commons

Different alleles at a single locus are represented as marbles (or as beads, or as jelly beans for the most voracious students of evolution). The metaphor is compelling: if each marble in the jar represents an allele possessed by an individual, then the whole jar represents the “gene pool”. By randomly sampling from the jar as a way of representing alleles being passed on to the next generation of offspring, we can see that in small populations the probability of losing an allele to chance is far greater than in larger populations.

But are alleles well-represented as marbles in a jar? This question — as absurd as it may seem — is at the heart of The “Random Genetic Drift” Fallacy‘s argument. Provine emphatically says “no”, suggesting that what we call genetic drift is really inbreeding, and inbreeding creates very different evolutionary outcomes than fixation of particular alleles at particular loci. There is no gene pool and there is no drift at particular gene loci because genes are located on chromosomes and chromosomes are replicated via the process of meiosis. Meiosis was not well-understood until just after Fisher, Wright, and Haldane had formulated the foundational theory of population genetics; according to Provine, population genetics has been out of synch with reality every since. In his own words:

Provine’s worry — expressed relentlessly throughout the book — is that by creating a theory that ignored the reality of chromosomes, the founding fathers of population genetics embedded a fatal flaw in their field.

Provine would like to see a better theory of genetic evolution on a chromosome. In particular he contends that genes go to fixation by inbreeding rather than by the mystical process of genetic drift. In smaller populations, the chances that individuals sharing recent ancestry will interbreed increases, and that increases the chances that the offspring of future generations will be homozygous not just for a particular allele but for all alleles found on a particular ancestral chromosome. Inbreeding is famously bad if it results in two copies of a deleterious allele coming together in the same individual’s genome, but inbreeding does not inevitably produce lower fitness individuals. Inbreeding can also purge variation that served as a barrier to adaptation in ancestral populations. Provine discusses selective breeding extensively, showing how inbreeding can sometimes allow a particularly valued phenotype to emerge.

What he is talking about is gene cluster evolution, a phenomenon that could explain how populations cross from adaptive peak to adaptive peak. Ironically, inbreeding — the phenomenon Wright failed to properly represent — could be a viable mechanism for change under Wright’s shifting balance theory. According to Provine, one of the reasons why Wright clung to the gene-centered view of evolution was that it produced “kaleidoscopic variation” in gene combinations, which would allow some new gene combinations to emerge at different adaptive peaks than the peaks occupied by ancestral populations. Wright believed that this was how populations crossed from lower to higher adaptive peaks across maladaptive valleys. But Provine suggests that inbreeding can also explain rapid adaptive shifts: so long as chromosomes maintain a variety of gene combinations that are only partially expressed in heterozygotes, the founding of a new small population and the inbreeding that follows such an event always has the potential to rapidly drive allele combinations to fixation. Most of the time such events would lead to the population collapsing, but as long as founding populations occasionally fixed allele combinations at new adaptive peaks, you don’t need a lot of recombination to produce new adaptations via small populations. Maybe what Wright perceived as critical to evolution — constant recombination of alleles on the chromosome — was in fact unnecessary.

A major strength of Provine’s argument rests on the fact that meiosis evolved. There is no glaring reason why having two copies of alleles at each gene locus would be evolutionarily optimal, or why those gene loci should be bundled in a particular order on a pair of analogous chromosomes, or why the independent assortment of chromosomes should occur when gametes are formed (in fact in bacteria this is not the way inheritance works, but diploid eukaryotic cells have had a pretty good run). Why meiosis evolved is not self-evident, but several viable adaptive explanations of meiosis have been forwarded. One hypothesis is that the chance process of meiosis is a way to prevent cheating by any particular gene; although some meiotic drive exists, for the most part meiosis is a non-biased process, which makes it hard for genes to replicate themselves at the expense of other genes. Another (potentially complementary) hypothesis is that meiosis creates the right balance between recombination (which generates novel gene combinations, grist for the evolutionary mill to produce future adaptation) and genetic stability (keeping alleles at different loci clustered on the same chromosome so that adaptive functions can be maintained). The problem with random genetic drift is that it ignores this potential balance, which might be responsible for the evolution of chromosomes and meiosis in the first place.

What about experiments? Certainly if random genetic drift has been an integral facet of population genetic theory, it should have been empirically demonstrated dozens of time by now, right? According to Provine, random genetic drift is a classic case of a theory that was so good that it did not merit testing:

He chronicles numerous experiments that claim to have produced allelic fixation via random genetic drift but which are clearly examples of inbreeding leading to loss of chromosomal diversity. Interestingly many of these studies were performed in Drosophila, which often maintain unusually high rates of recombination.

Provine also wants to make it clear that the shadow of pop-gen’s founding fathers is long. He devotes a chapter to Kimura and Ohta, the founding father and mother of neutral theory, showing that Kimura’s contribution was to extend random genetic drift thinking by applying Kolmogorov‘s diffusion equations to the problem of gene fixation. Like his predecessors, Kimura treated each gene locus as if it existed on its very own chromosome. Again, this over-simplification allowed the application of models that would not otherwise apply to a more nuanced depiction of genetic inheritance. In 1975 Kimura and Ohta published a paper in which they explained their rationale by insisting:

Provine suggests that Kimura’s neutral theory is the updated foundation of population genetics, and like its predecessor is a shaky and unrealistic platform on which to base our understanding of genetic evolution.

If Provine is right about the flaws of population genetics, we have some serious problems with the way that we depict the processes of evolutionary change.

What if this book is right?

Although the marble metaphor is meant to explain how a gene goes to fixation via random processes, it is also the metaphor that is at the heart of all traditional population genetics thinking. For genes to be the target of natural selective processes, they must be independent of other genes. If our nuclei were like a bag of marbles filled with all of our genes, each an individual particle that could be passed on to our offspring, then perhaps each gene could be considered an independently-evolving entity. Of course then the whole concept of a locus would not make a lot of sense, as the point of thinking about a locus is that it is a particular gene that performs a particular function by virtue of the fact that it exists in a particular place on the chromosome. So is it strange to think of genes as independently-evolving entities when they spend all their time existing on a chromosome with lots of other genes?

Richard Dawkins’ The Selfish Gene (1976) has always been an important book to me. When I first read it, it was a model for how to communicate science to the layperson. When I returned to it later, it was a model for understanding how the evolutionary process worked. As I have aged as an evolutionary biologist, it has come to symbolize the danger of theory accepted as reality and metaphor accepted as mechanism. In the cultural evolution of a scientific field, some ideas have the potential to propel the field forward; others have the potential to lead the field on an extended labyrinthine voyage into a dead end. As I have come to better understand its argument and impact, I am pretty sure that The Selfish Gene fits into the latter category.

The foundational argument of The Selfish Gene is that natural selection acts on individual gene loci (in other words, Dawkins is a disciple of Fisher, Wright, and Haldane). If the assumption that genes evolve independently is wrong, most of what the rest of The Selfish Gene has to say is wrong. Dawkins knows this, and goes to great pains to justify that selection on genes — rather than on chromosomes or on individuals or on groups — is what drives the evolutionary process. He introduces the idea of recombination early in the book and then — in true Dawkins style — helps his reader understand the effects of recombination with a clever metaphor.

We are asked to think of each gene locus on a chromosome as a seat on a rowing scull that optimally will be occupied by a strong rower (or perhaps a coxswain with a loud voice and good rhythm). Via this metaphor Dawkins concedes that genes exist on a chromosome and therefore do not act in isolation. But he asks the reader to imagine that although a given race in a given boat (representing a set of gene variants in a particular individual organism) will depend on the combination of rowers, such “races” in the biological world are conducted with a great variety of rower combinations. More fit rowers will do better on average in each of their races, even though they find themselves with different rowers each time they race. If these racers were stuck with the same rowing partners during each race, the best rowers at each position on the boat might not be victorious overall. But because recombination of genes on the chromosomes mixes up gene combinations, the best metaphor is a rowing tournament that picks the best rower at each boat position by randomly forming rowing teams over the course of numerous races. The best rowers on average win the most races and are “selected” for their particular position in the scull.

There are a lot of weird elements to Dawkins’ metaphor, but in the context of Provine’s argument I want to focus on one issue: does recombination really lead to independently-evolving genes at each locus on a chromosome? That is the assumption of early models of genetic drift, and the assumption that Dawkins defends in order to support his gene-centered model of evolution. Clearly Provine does not agree, suggesting that recombination cannot happen fast enough for us to consider each gene locus as an independently-evolving entity. Provine is concerned about the idea that a particular neutral allele can go to fixation independently, but if recombination is not sufficient to allow independent drift of alleles, it is also unlikely to allow for selection to act on individual alleles at particular gene loci. And that would mean that it’s not just our models of neutral evolution that are problematic: the way that population genetics models depict natural selection would also be wrong.

What made the modern evolutionary synthesis important was that it plugged a serious hole in Darwin’s theory of evolution. Darwin had wonderfully-prescient ideas about how evolution proceeded by modification with descent, but he lacked a real model for descent. Mendelian genetics provided that model, and Fisher, Wright, and Haldane provided the mathematics that would bridge our understanding of genetic inheritance with Darwin’s idea of natural selection. If everything about those mathematics misrepresented the realities of genetic inheritance — which is the crux of Provine’s argument — then there never was a synthesis between Darwin’s theories and modern genetics. Yikes, that would be a scary state of affairs.

This book makes it clear how problematic it is that population genetics had just a few founding fathers. The early influence of these scholars blazed an influential path, but the question is whether that path has led us astray. What if our science is dogmatic, sticking to early ideas even when they don’t make sense? What if our science is lazy, failing to tackle harder problems when easy (false) solutions are available? These are the frightening questions that The “Random Genetic Drift” Fallacy provokes.

Is this book right?

Reading this book made me realize that I do not have a strong enough foundation in population genetics to really assess the validity of Provine’s argument. I am sorry for leading you along so far, just to say I don’t know. But I don’t, and that’s because I am not familiar enough with the population genetics literature to determine whether Provine’s depiction of the field is accurate. If his depiction is accurate then we have some serious cleaning up of population genetics to do, but I have a sneaking suspicion that Provine’s depiction of the modern state of population genetics might be a bit incomplete.

Provine only briefly discusses linkage disequilibrium, but it is a critical concept in understanding his argument. If small populations are what lead to fixation of an allele by random processes, as that allele rushes towards fixation there is a kind of race between recombination and inbreeding effects. Inbreeding effects in small populations have the prospect to cause whole chromosomes to go to fixation in the population, dragging all the alleles on that chromosome to fixation as well. Meanwhile, recombination breaks up existing chromosome combinations, making it possible for particular loci to go to fixation while other loci on the same chromosome maintain standing variation in the population. The big question is which of these processes happens faster? If the effects of inbreeding are rapid — as is potentially the case in Provine’s examples from plant breeding — then whole chromosomes will go to fixation. But if recombination happens rapidly, then the only way fixation can happen is via genetic drift at particular loci. Whether the traditional representation of neutral evolution is accurate really depends on how rampant inbreeding is and how frequently recombination occurs.

Does contemporary population genetics ignore the reality of meiosis and chromosomes as Provine implies? From what I can tell, the answer is probably no. That is to say that Provine may be overstating the disaster wreaked by the random genetic drift fallacy. This month I have been doing a lot of research on recent human evolution, and pretty much every paper that I have read is focused on the issue of recombination, specifically how long it takes for recombination to break up the very genetic hitchhiking at non-selected loci that Provine wants us to keep our eyes on. Ironically, looking at the decay of linkage disequilibrium is one of the best ways to detect recent selection in species like humans, who have low recombination rates and probably have not evolved by the classic selective sweeps envisioned by early population genetic theories (Hernandez et al. 2011, Fu & Akey 2013). So the truth about population genetics is a bit more nuanced: while scientists are still looking for evidence of selection within the background of neutral evolution, they do so by acknowledging that genetic processes are far more complex than the simple gene-centered models of Fisher, Wright, and Haldane.

It is also now widely acknowledged that early studies of humans showing strong selection on particular genes represents “low hanging fruit” (Scheinfeldt & Tishkoff 2013). Most of our important phenotypes emerge from the complex interactions of many genes, and detecting selection for these phenotypes is a lot more difficult (Pritchard 2010). Ironically, this may mean that the modern evolutionary synthesis is incomplete for slightly different reasons than those proposed by Provine in The “Random Genetic Drift” Fallacy. Our problem is that we treat genes in isolation, and that we need a population genetic theory that acknowledges not just individual chromosomes but the nature of the entire genome. But misunderstanding how drift works is really just the tip of the iceberg: the bigger problem is misunderstanding how selection works. Based on what I have read, there is not a coherent population genetics theory explaining how polygenic traits evolve. If polygenic traits are adaptively important, that means that we have not really completed the synthesis of Darwinian logic with genetic realities.

While I cannot provide a thoroughly-informed assessment of whether Provine is correct in suggesting that random genetic drift has torpedoed the entire population genetics endeavor, I can see that his critique needs to be applied to the way we teach about population genetics in evolutionary biology courses. Personally, after reading this book I am done with the marble metaphor; although it is a valuable way of getting students to think about how stochasticity can impact the evolutionary process, thinking of alleles as marbles of different colors being picked out of a bowl seriously misleads students about how genetics actually work. And if you want to prime your students for being able to think about the complexities of genetic architecture, the marble metaphor is a disastrous way of depicting random genetic change. In an era where the scientific challenge in both genetics and evolution is to understand how networks of genes produce phenotypes (Weiss & Buchannan 2009), students are mis-served by the single-gene thinking of this lazy classroom metaphor derived from a lazy treatment of genetics. Rather than delivering students into the arms of selfish gene thinking, I intend to talk about inbreeding and loss of chromosomes as one way of understanding why small populations are prone to fixing particular alleles despite those alleles having no selective advantage. Yeah, that’s a messier idea to have to communicate, but it turns out that genetics is far messier than the early population genetics theorists chose to depict.

Will this book have a lasting impact?

There are many charms that this book lacks. Primarily, its main message is bluntly hammered over and over again until one wants to just tell the book hey, I got it. The writing is often inelegant and not particularly reader-friendly. The book can be a bit difficult to follow, especially towards the end where its coherence begins to decline rather precipitously. It is kind of a glorious mess, and you have to be willing to wade through the mess to see the glory. This book needed an editor, but on the merits of its ideas it also deserved an editor. I can’t help but love the roguish, messy, passionate nature of what Provine has written here, despite the many faults of the text.

There are a lot of questions about this book that I wish that I had asked Dr. Provine before he died. Certainly one of these would be why did you self-publish this book? Was it that you wanted to make sure that the book was released as soon as possible? (Academic books are published at a notoriously geological pace). Was it that no publisher would support the controversial ideas presented in this book? (Science publishing can be pretty conservative, resisting the propagation of ideas that contradict the foundational orthodoxies of a field). Perhaps the answers to these questions do not matter, as we have the self-published book to speak for itself. But I cannot help but wonder how much more effective this book could have been had it been properly edited, as would have happened if its publication had been supported by an academic press. There are some important ideas to keep alive in this book, and I worry that a combination of obscurity (because this book is self-published) and lack of elegance (because this book needs a good editing) will prevent The “Random Genetic Drift” Fallacy from being as widely read as it deserves.