Substack cometh, and lo it is good. (Pricing)

Misconceptions in evolutionary biology

Chad Orzel is asking about misconceptions in science that irritate. Evolgen and Afarensis have chimed in. My problem is not an misconception, it is a pet peeve. As I’ve noted before, random genetic drift is a catchall explanation for everything.
I am not saying drift is not powerful, it is the basis for the neutral theory of molecular evolution. This theory states that the rate of substitution on a neutral locus is proportional to the rate of mutation. Substitution would be when you have allele X at 99% frequency at time 1 and allele Y at 99% frequency at time 2 on a particular locus. Y has substituted for X on the locus over time. By “neutral” we mean that there is no fitness deviation for the mutation, Y is not more fit than X. So why would Y substitute for X? Because over time sampling processes result in imperfect representation of the genetic frequencies in generation n in generation n + 1,1 shift the proportions every generation, and given enough time all alleles will be substituted by another variant, assuming that the substituted allele is neutral relative to the ancestral variant. The insight of neutral theory is that this rate of substitution is dependent only on the rate of mutation, it doesn’t depend on population size. Verbally, recall that the number of mutations bubbling within a population is proportional to the size of the population, while the probability of fixation is inversely proportional to the size of the population because random genetic drift is far more powerful in small populations. More formally (but leaving out Greek symbology):
rate of substitution = probability of fixation X number of mutations
number of mutations = rate of mutations X population size
probability of fixation is the 1/(population size)
ergo, when you multiply the equation out only rate of mutation remains as population size cancels out
Neutral theory was prompted in part by the Lewontin and Hubby paper which I have pointed to before, which showed that the extent of polymorphism, genetic diversity, was far higher than classical evolutionary genetics predicted. The older schools, working from a priori, assumed that mutation would generally be purged from the genetic background, so that the genome would be constrained toward the fitness optimal state. It is probably true (depending on the organism and how you define genome) that most mutations are purged, but, of those that aren’t, neutral theory implies that most have no fitness implication, rather than being favorable. So why does this imply polymorphism? Because it takes time for mutants to be fixed, rough 4 X the breeding population size in terms of generations (4Ne), so the transition phases will exhibit polymorphism (in a large population across numerous loci there will always be some portions of the genome in transition rather than being fixed). Large populations will be characterized by a great deal of polymorphism, while small populations less so (mutants show up infrequenty, and become fixed quickly if they do substitute on a locus).
This review is to leave no doubt that I do think random genetic drift is important conceptually. I am not Will Provine. But, it is called the neutral theory of molecular evolution for a reason: it does not imply we will see a great deal of gross phenotypic variation because of random walk processes only (selection plays the role of directional organizer on the mutational variation). Neutral theory is a good null hypothesis for molecular evolution, but, that does not mean that we should assume that a newly evolved trait is due to random genetic drift if we can’t pin down the exact precise functional role it plays in a new adaptive niche, or its fitness implications. Consider adult lactose digestion, we know why that evolved and rose in frequency over the past 10,000 years in some populations (Europeans, other groups in Western Eurasia and among some African groups), but I have seen little discussion as to what cost this phenotype exacts. I have even heard that there is recent data coming out of the HapMap that East Asians might have evolved a reverse form which heightens lactose indigestability as adults! In the ear wax paper the authors give some weak selective hypotheses (at least I thought they were weak), but, I think the LDE signatures and the distribution of the alleles is strong evidence that selection is at work, we just don’t know what.
Post rant note: On selection vs. drift, consider that the probability of fixation for a new mutation is 1/2N if it is neutral, that is, not good, not bad. So, if the breeding population is 10, that means a new mutant as a 5% chance of fixation, and it should take about 40 generations (4Ne). Now, consider an alternative scenario of an infinite sized population, the probability of fixation for a new mutant that is benefit is 2s, s being the selection coefficient which measures that allele’s relative fitness vis-a-vi the population mean. If s is 0.02, you have a 4% chance of fixation. Since the population size is infinite you have don’t have to worry about drift since 1 over infinite ~ 0. The point is that drift vs. selection is about population sizes. If you look at the ear wax map you see a clinal gradation, and it is just implausible that drift should have worked over enormous swaths of East Asia 10,000 years ago (the derived allele is in the New World, that gives you a lower time bound). You can imagine scenarios where a small population fixes, and then somehow expands, but with the LDE data I don’t think that’s more parsimonious than selection. If drift is working separately in the all the populations, than you shouldn’t see a geographical cline on the continental level, stochastic processes are stochastic.
Update: John comments.
1 – Going back to the binomial distribution, which is so easy it is sweet, if you have N as the number of individuals in a population, and they are haploid (one copy of a gene), and the frequency for alleles A & B are both is .5 , and assume that all subsequent generations have the same number of individuals, then the expected deviation due to sampling variance would be inversely proportional population. Variance of the proportion is going to be pq/N, so, as the number of individuals increases the variance of the proportion between generations decreases. See here for the real math, either more intelligible or more opaque depending on your taste.

Posted in Uncategorized

Comments are closed.