Chance and Selection

Five real cases

Examples of genetic drift

Five documented cases, from a laboratory experiment to a seal colony, each with a figure that runs the same random sampling.

5 min read · 5 interactive figures

Genetic drift in its simplest form: twelve beetles, no selection, each generation a random draw of 24 gene copies from the last. Press New population a few times: sometimes orange wins, sometimes blue, for no reason at all.
On this page
  1. 1. Buri’s 107 bottles of fruit flies
  2. 2. Most new mutations are lost
  3. 3. The Old Order Amish: a founder effect
  4. 4. Pingelap and colour blindness
  5. 5. Northern elephant seals: a bottleneck you can measure
  6. What the examples have in common

Genetic drift is change in how common a gene variant is caused by chance: which individuals happen to survive and breed, and which of their two gene copies each offspring happens to receive. It has no direction, and it is strongest in small populations. Good examples are those where the chance can be measured, or where its fingerprints are unmistakable. Here are five, each with a figure that runs the same exact random sampling. Chapter 3 gives the theory.

1. Buri’s 107 bottles of fruit flies

The cleanest example is an experiment. In 1956 Peter Buri set up 107 bottles of Drosophila melanogaster, each founded by 16 flies carrying exactly 16 copies of the eye-colour allele bw75 among their 32 gene copies, and each generation moved 8 males and 8 females chosen at random to a fresh bottle[1]. Every bottle was the same at the start and nothing favoured either allele. After 19 generations, 30 bottles had lost bw75 and 28 had fixed it, while the average frequency over all 107 stayed close to one half. Directionless sampling, run 107 times, produced populations at both extremes[2]. established

Figure 1 Buri’s 107 populations, generation by generation (bars), against the exact Wright–Fisher prediction for the effective size you choose. Press Play. Notice the bell flattening into a U as bottles drift to loss or fixation; with the census size of 16 the model is too slow, and about 9 fits.#

The flies also show why biologists talk about effective size. The bottles drifted as if each held about nine flies, not sixteen, because some flies left many offspring and others none[3].

2. Most new mutations are lost

Every new mutation starts as a single copy, and a single copy is at the mercy of the lottery. In the neutral case its chance of ever taking over is 1/2N, so in a population of a million, one in two million; nearly all are lost within a few generations by chance alone[4][3]. Even a beneficial mutation that gives its carriers a 1 % advantage (s = 0.02 with h = ½, in this site’s convention) is lost about 98 times in 100. This is drift at work in every population, every generation. established

Figure 2 Two hundred populations of N individuals, each starting with one copy of a new mutation; AA has fitness 1 + s and the carriers 1 + s/2. Run them: nearly all tubes stay blue, because the single copy is lost within a few generations. Set s to zero for a neutral mutation.#

3. The Old Order Amish: a founder effect

Ellis–van Creveld syndrome, a recessive form of short-limbed dwarfism, is rare worldwide (1 in 60,000 to 200,000 births) but much more common among the Old Order Amish of Lancaster County[5]. Its high frequency is limited to a single Amish community, whereas cartilage–hair hypoplasia, another recessive condition, is frequent in at least three; one community fits one carrier among that community’s founders[6]. A founder effect is drift concentrated in one generation: a few people carry a sample of the source population’s genes, and a sample of a few can be very unrepresentative. documented

Figure 3 Ten founders leave a population in which an allele has frequency 1 %. The bars give the exact chance of each founder frequency. Notice that most boatloads carry none, but almost one in five carries it at 5 % or more, five times the source frequency. Draw a few boatloads.#

4. Pingelap and colour blindness

On Pingelap, a Micronesian atoll, complete achromatopsia (no colour vision at all) is common; elsewhere it is rare. The mutation is in the gene CNGB3[7]. The accepted history is a bottleneck: a typhoon in the late eighteenth century left few survivors, at least one of them a carrier, and the population that regrew from them carried the allele at a high frequency[8][9]. reconstructed

5. Northern elephant seals: a bottleneck you can measure

Northern elephant seals were hunted to perhaps 20 animals by 1892 and have since recovered to more than 220,000. Comparing museum genomes from before the bottleneck with modern ones, Hoelzel and colleagues found that average heterozygosity fell from 0.00142 to 0.000176: the species kept about one eighth of its genetic variation despite a more than ten-thousandfold recovery in numbers[10]. Drift during the crash removed variation, and recovery did not bring it back. established

Figure 4 A population of N₀ crashes to Nb for a few generations, then recovers. Drag the size during the crash in the top panel, or use the sliders. Notice that heterozygosity (solid) falls during the crash and stays down after the numbers recover; the dashed line is the formula.#

What the examples have in common

In every case the cause is sampling: a finite number of gene copies is passed on, and which ones is partly luck. The effect is largest where the number is smallest (sixteen flies, a boatload of founders, the survivors of a typhoon or a hunt), and its signatures are the same each time: rare variants lost, a few made common by accident, and populations that started alike ending up different. For the difference between the founder and bottleneck cases, see founder effect vs bottleneck effect; for drift set against selection, Chapter 6.