MLB7052.LT: Modern Methods in Molecular Ecology
TalTech / UT Institute of Technology
Loeng 10 — populatsioonigeneetika (population genetics).
Every method in this course produces genotypes. This lecture is about what you are allowed to conclude from them — how allele frequencies behave when nothing is happening, and how each thing that is happening leaves its own signature in the numbers.
Population genetics is where molecular ecology stops being descriptive.
A marker on its own tells you that two individuals differ. Frequencies across a population, compared against an explicit null model, tell you why.
Everything in this lecture is one idea repeated: write down what the frequencies would do if nothing were acting, then read the departure.
A population (populatsioon) is a group of individuals of one species, freely interbreeding, occupying a defined territory and separated from other groups of the same species by some form of isolation.
Märkus
The set of all gene variants across the genotypes of all individuals in a population is the gene pool — geenifond, genofond.
1970 — Richard Levins: “populations of populations”.
Metapopulations exist where a population is divided into subpopulations whose dynamics are set by the balance between local extinction and recolonisation.
Divided into core and satellite populations, and also:
A deme (deem) is a local, isolated subpopulation — an interbreeding group of related individuals that is unstable and short-lived (1–3 generations, no independent evolution of its own).
Polar bear subpopulations and their trends (WWF Canada). Conservation status is assigned per subpopulation — which only means something if the subpopulations are genetically real.
Population genetics studies change in the gene pool: the origin of variation, the transmission of genes, the inheritance of traits from generation to generation, and the effect of natural selection on the genetic structure of a population — the regularities that hold for the population as a whole.
The British mathematician Godfrey H. Hardy (1877–1947) and the German physician Wilhelm Weinberg (1862–1939) published independently in 1908, describing the relationship between allele and genotype frequencies in populations.
It lets you predict genotype frequencies in the next generation from the allele frequencies of the parental generation — under the assumptions of an “ideal” population.
In a panmictic population (tasakaaluline e. panmiktiline populatsioon):
W. Weinberg
G. H. Hardy
For a single autosomal, diallelic gene:
where is the frequency of AA, of Aa, and of aa.
Sex-linked genes (suguliitelised geenid):
One autosomal, polyallelic gene — the ABO blood groups have three main alleles (, , ); and are codominant with each other and dominant over . For three alleles:
30 individuals scored on a gel: 10 PP, 13 PQ, 7 QQ.
, so .
Expected under HW, out of 30:
A test asks whether expected and observed differ significantly:
.
at 1 df gives : the null hypothesis cannot be rejected, and this sample is in HW equilibrium.
Natural populations are, to a greater or lesser degree, changing and out of equilibrium. The causes are, above all:
Tähtis
Each of these four is a deviation from an assumption of HW — which is exactly why HW is worth stating. The rest of this lecture takes them one at a time.
Effective population size () refers to the number of individuals in a population who actually contribute offspring to the next generation. It is an idealised number reflecting genetic diversity, not a headcount.
Unlike the actual population size (), accounts for factors that reduce genetic diversity:
populatoR-Aa
Everything from here on can be simulated in the course app:
The app runs replicate populations forward under the controls you set:
And it plots exactly the quantities defined in this lecture:
· · · · ·
Start here: tick Infinite population (no drift), leave migration, mutation and the fitnesses neutral, and watch stay flat. That flat line is Hardy–Weinberg. Every later section is a way of bending it.
Genetic drift (geenitriiv; genetic drift, random genetic drift, allelic drift) is the change in allele frequencies across successive generations caused by random sampling.
A population does not form all possible mating pairs with equal probability — and the smaller the population, the more sensitive it is.
Special cases:
Same starting frequency, same neutral alleles, no selection whatsoever. Only the sample size differs. In the small populations alleles reach fixation or loss within a century; in the large ones they wander and persist.
Each line is one replicate population, , binomial sampling of gametes per generation — the same process populatoR-Aa runs interactively.
Gene flow (geenivool, geenisiire) is the exchange of genetic material between populations, or between parts of a population, through the migration and interbreeding of individuals — or, in plants, the movement of pollen into another population.
Recent separation with no gene flow, versus old separation with continuous low gene flow — two histories that can leave similar distances.
Gene flow is the movement of alleles, mediated by individuals or their gametes, between populations.
Allele frequency data permit inferences of gene flow, ultimately based on how much migration disturbs Hardy–Weinberg expectations. F-statistics — particularly — provide the basis of most approaches.
, the average number of successfully reproducing migrants per generation, is classically derived from :
on the left for diploid nuclear genes, on the right for uniparentally inherited haploid genomes such as mtDNA.
is equivalent to — and one migrant per generation is the classic threshold above which subpopulations will not diverge by drift alone.
It is theoretically independent of population size, but simulations show this is not strictly true; in practice more than one migrant per generation is usually needed to prevent inbreeding in natural populations (Vucetich & Waite 2001).
Hoiatus
That formula assumes an island model — movement between all subpopulations equally probable, distance irrelevant, and equilibrium between migration and drift. Real landscapes rarely oblige.
In a small population, allele frequencies change through random drift — and with them the frequency of heterozygotes, which is what population heterozygosity expresses.
The equilibrium heterozygosity under a balance of mutation and drift depends on the effective population size. The mutation rate is of the order of .
Mutation pressure raises variability, drift lowers it; selection removes deleterious recessives. Heterozygosity settles where these balance.
Common ash, Fraxinus excelsior (harilik saar)
Papers typically report several measures of genetic diversity, and these usually include and . Comparing the two tells you whether heterozygosity within a population differs significantly from the HW expectation — here, across three ash populations at five microsatellite loci (Moran et al. 2002).
Märkus
Look at Population 1, locus 1: but . A large heterozygote deficit — inbreeding, a null allele, or unrecognised substructure. The comparison is the diagnostic.
— inbreeding coefficient of an Individual relative to the Subpopulation:
— fixation index of a Subpopulation relative to the Total population:
— inbreeding coefficient of an Individual relative to the Total population: overall inbreeding.
= average observed heterozygosity across subpopulations; = average expected heterozygosity across subpopulations; = expected heterozygosity of the total population.
Two subpopulations, each with at 0.8 and at 0.2, each in HW proportions. , so
No inbreeding, and no structure — the subdivision is a line on a map, not a genetic boundary.
Situation 2
We worked Situation 2 through in the lecture. The version printed in the book is wrong — use the lecture calculation, not the table.
Gene flow and genetic drift are two fundamental processes influencing genetic variation across populations, and they pull in opposite directions:
Wright’s equation relates to gene flow, where is the effective population size and the migration rate:
Tähtis
alone cannot separate “much migration” from “recent separation”. Both give a small number. You need an independent handle on time.
In practice, researchers use pairwise values against geographic distance to visualise and analyse genetic structure.
Shown here among collared lizard populations:
Varying illustrates different scales of population structure.
Clustering methods are popular, but open to misinterpretation: for non-simulated data there is never a “true” value of , only an approximation that is useful for a given question.
Differentiated demes of Atlantic salmon Salmo salar (Garant et al. 2000).
Hierarchical analysis of molecular variance (AMOVA). Note where the variance actually sits: 96.6% within samples — and yet the among-year and among-site components are the significant ones.
The idealised population
Random mating, simultaneous birth of each new generation, constant population size, equal numbers of children per parent. Generally: is the number of individuals an idealised population would need for some specified quantity of interest to match the real population.
A population bottleneck is a radical reduction in population size, usually accompanied by a loss of genetic diversity.
Northern hairy-nosed wombat (põhja-karvikvombat): -obs exceeds -exp at eight of nine microsatellite loci.
More mutations occur in total per unit time in a large population than in a small one.
The average time between the mutation and the fixation of a new allele is proportional to .
Shown: (a) the expected relationship between and , and (b) the actual relationship for a range of animal species.
is defined as the number of individuals in an idealised population that has a value of any given population genetic quantity equal to the value of that quantity in the population of interest. Which quantity you pick gives you a different .
The two quantities identified by Wright:
Today is usually estimated empirically with respect to sojourn or coalescence time, as within-species genetic diversity divided by the mutation rate — the coalescent effective population size, the one used in conservation ecology. It has been extended to overlapping generations and structured populations.
Tähtis
is affected by sex ratios and variance in reproductive success, and over time it varies as a harmonic average over multiple generations — which is why one bad year dominates.
When the sex ratio departs from the Fisherian 1:1, effective population size is given by
with the number of males and the number of females.
With 80 males and 20 females:
100 adult zebras: 60 females, 12 males with harems, 28 bachelor males.
So even with 100 animals, this is equivalent to 20 males and 20 females in an ideal, randomly mating population.
Grizzly bear, Yellowstone National Park
Suppose the abundance of grizzly bears is , with age structure data:
, the number born every year that survive to reproductive age:
Adjusting for generation time:
— only about a quarter of the total bears.
Population size varies over time. For non-overlapping generations, effective population size is given by the harmonic mean of the population sizes — the reciprocal of the arithmetic mean of the reciprocals:
For = 10, 100, 50, 80, 20, 500 over six generations:
The arithmetic mean is 126.7. The harmonic mean is 30.8.
Tähtis
The smallest generation dominates. A single crash to sets the genetic fate of the lineage, no matter how good the other five years were.
If population size is to remain constant, each individual must contribute on average two gametes to the next generation.
An idealised population assumes this follows a Poisson distribution, so the variance in the number of gametes contributed, , equals the mean, i.e. .
In natural populations the variance is often larger: the vast majority of individuals may have no offspring at all, and the next generation stems from a small number of individuals, so .
If the variance of is less than 2, is greater than .
With for dioeciousness (having separate females and males):
Maximum likelihood and coalescent-based approaches using repeat sampling estimate as the size of the population that best explains the variance in allele frequencies observed between samplings.
In theory can also be estimated from its expected relationship to heterozygosity:
where is the mutation rate. But obtained this way can rarely be attached to specific isolated populations, because their diversity usually includes alleles that arose before the current structure — it averages over long time periods, and is more appropriate for entire species over evolutionary time.
Temporal method (Waples 1989) — based on the expectation that drift increases as decreases, measuring temporal change in allele frequencies between generations. It requires sampling a population on two successive occasions:
is a statistic reflecting the amount of drift between the two sampling times; is the time in generations between samplings, and , the sample sizes in the two generations, for organisms with non-overlapping generations.
If matings in a population are not random, the population does not obey the HW law. Two types of non-random mating are distinguished:
Consanguineous and assortative mating both raise the level of homozygosity and reduce the number of heterozygotes:
— the inbreeding coefficient — is the selection effect from mating between individuals of similar genotype and shared ancestry. means no inbreeding; is complete inbreeding.
Märkus
Note what inbreeding does not do: it does not change and . It only repackages the same alleles into more homozygotes — which is why it shows up as a heterozygote deficit against , exactly the signature seen in the ash data.
Absolute fitness () against number of generations (). is the proportional change in the abundance of a genotype over one generation attributable to selection.
Traits are usually affected by the alleles of many genes. Fitness then measures quantitative traits — body size and weight, disease susceptibility, fecundity and so on.
Selection acts on quantitative traits in three ways:
Simplified, fitness can be determined at the level of the gene:
The fitness (kohasus, ) of individuals — the ability to survive and produce viable offspring — is a quantitative trait.
The mean fitness of the population is:
growing, stable, declining population.
Dobrynin et al. 2015
Open it and work through these
1 · Hardy–Weinberg. Tick Infinite population (no drift), set all three fitnesses equal, migration and mutation to zero. Plot . Nothing moves — that is the null model.
2 · Drift. Untick it. Set population size to 25 and run 20 populations for 100 generations. Then set size to 500. Compare how many replicates hit fixation.
3 · Bottleneck. Small population size, many replicates: watch fall and see how variable the outcome is between runs.
4 · Gene flow. Several populations, then raise the migration rate from 0. Plot — check the collapse against .
5 · Structure. Plot and together. Watch grow as the gap opens between them.
6 · Selection. Set the fitness of BB below AA and AB and plot alongside . Then make the heterozygote AB the fittest and see what stops from reaching 1.
Tähtis
Every quantity the app plots — , , , , , — is defined somewhere in this lecture. If one of them is not yet meaningful to you, that is the section to go back to.
Because it is a null model, not a description. It states exactly what the genotype frequencies would be if nothing were acting, so any significant departure is evidence that something specific — selection, drift, inbreeding, migration or hidden structure — is acting, and the direction of the departure narrows down which.
Inbreeding, a null allele at that locus, or unrecognised substructure within the sample (a Wahlund effect). Inbreeding and substructure would affect most loci; a deficit at a single locus points to the marker itself, or to selection at a linked site.
Because a low is equally consistent with a recent common origin and no gene flow at all. measures how far apart the frequencies are, not how they got there, and the island-model conversion to assumes an equilibrium between migration and drift that a recently separated pair has not reached.
, for anything about genetic diversity, inbreeding or the rate of drift — it is the size the population behaves as, genetically. The census size still matters for demography and extinction risk. They answer different questions and quoting one for the other is the common error.
Because effective size across generations is the harmonic mean, which is dominated by the smallest terms — here 30.8. Genetic diversity lost in the crash generation is not restored by the good years that follow; drift is not averaged, it is accumulated.
Tähtis
peeter.laas@taltech.ee · peeter.laas@ut.ee
Try the simulator:
Also worth your time:
Fall 2026