Advancedadvanced biology

Population Genetics: The Mathematics of Allele Frequencies

Learning objectives

  • Derive Hardy-Weinberg genotype frequencies and state the equilibrium assumptions.
  • Quantify selection with fitnesses and selection coefficients, and predict allele frequency change.
  • Explain genetic drift through effective population size and predict heterozygosity loss.
  • Analyze mutation-selection balance and migration-selection balance.
  • Interpret Fst and linkage disequilibrium as measures of population structure and association.

Hardy-Weinberg: the null model of evolution

In 1908, G. H. Hardy and Wilhelm Weinberg independently showed that in a large, randomly mating population with no selection, mutation, migration, or drift, allele frequencies determine genotype frequencies as p^2 + 2pq + q^2 = 1, and these frequencies remain constant generation after generation. The result is profound: Mendelian inheritance alone does not change allele frequencies -- something else must drive evolution. The equilibrium rests on idealized assumptions, which is precisely its power: deviations reveal the forces at work. Excess homozygotes signal inbreeding or population subdivision (the Wahlund effect); deficiency of heterozygotes at a locus under balancing selection; systematic shifts, directional selection. The chi-square test for Hardy-Weinberg equilibrium is therefore the workhorse of quality control in genetic studies -- in genome-wide association studies, SNPs violating it are often genotyping errors. Like the ideal gas law, Hardy-Weinberg is a null model whose violations are the interesting biology.

Selection in equations: fitness and the selection coefficient

Selection is quantified through fitnesses: assign genotypes AA, Aa, aa relative fitnesses w_AA, w_Aa, w_aa, often parameterized with a selection coefficient s and dominance coefficient h (w_AA = 1, w_Aa = 1 + hs, w_aa = 1 + s). Mean fitness w-bar is the frequency-weighted average, and the change in allele frequency per generation is delta-p = pq[p(w_AA - w_Aa) + q(w_Aa - w_aa)] / w-bar. For directional selection on a favored allele, this simplifies: with complete dominance, delta-p is approximately spq^2 / w-bar, revealing why rare recessive beneficial alleles spread agonizingly slowly -- they hide in heterozygotes. Haldane's classic result showed a new beneficial mutation's fixation probability is about 2s, meaning most advantageous mutations are lost to drift when rare. Balancing selection maintains polymorphism: heterozygote advantage (sickle-cell trait against malaria) and frequency-dependent selection produce stable equilibria where delta-p = 0 at intermediate frequencies. These equations turn verbal arguments about adaptation into testable predictions.

Drift: sampling error as an evolutionary force

In finite populations, allele frequencies fluctuate by chance -- genetic drift, the Wright-Fisher model's core. Each generation samples 2N gene copies from the parental pool, so allele frequency variance per generation is p(1-p)/(2N). Small populations drift fast: the critical parameter is the effective population size Ne, usually much smaller than census size because of unequal sex ratios, variance in reproductive success, and population bottlenecks. A neutral allele's fixation probability equals its current frequency, and heterozygosity decays as H_t = H_0 (1 - 1/(2Ne))^t -- in a population of Ne = 50, half the heterozygosity vanishes in about 69 generations. Drift dominates selection when s is much less than 1/(2Ne): in small populations, even beneficial alleles behave neutrally. Founder effects and bottlenecks -- the cheetah's near-zero diversity, Ashkenazi Jewish disease alleles -- are drift's signatures. Conservation genetics is largely applied drift theory: maintaining Ne above a few hundred is the quantitative basis for preserving evolutionary potential.

Mutation, migration, and balance

Mutation is evolution's ultimate source of variation but a weak directional force: with mutation rate mu around 10^-8 per base per generation, mutation alone would take geological time to shift frequencies. Its importance emerges in balance with stronger forces. Mutation-selection balance predicts the equilibrium frequency of deleterious alleles: for a recessive lethal, q-hat is approximately sqrt(mu/s); for a partially dominant one, q-hat is approximately mu/(h*s). This explains why severe recessive diseases like cystic fibrosis persist at low frequencies -- selection cannot purge alleles hidden in heterozygotes faster than mutation recreates them. Migration homogenizes: with migration rate m per generation, allele frequency differences between populations decay geometrically, and the equilibrium under migration-selection balance depends on the ratio m/s. When migration overwhelms selection (m >> s), local adaptation collapses -- the swamping effect central to debates about gene flow and speciation. These balance equations show evolution as a tug-of-war, with equilibria set by the relative strengths of opposing forces.

Structure and association: Fst and linkage disequilibrium

Wright's Fst measures population differentiation: Fst = (H_T - H_S)/H_T, the fraction of total heterozygosity attributable to between-population differences. Values near zero mean panmixia; human continental populations show Fst around 0.1-0.15, modest differentiation reflecting recent common ancestry. High Fst at a locus flags possible local adaptation, though demographic history confounds naive scans. Linkage disequilibrium (LD) -- nonrandom association of alleles at different loci, measured by D or r^2 -- decays with recombination and underpins association mapping: a marker SNP tags a causal variant only if LD links them. LD blocks reflect recombination hotspots, bottlenecks, and selection; the human HapMap and 1000 Genomes projects mapped LD genome-wide, enabling GWAS. Coalescent theory completes the picture, modeling how gene copies trace back to common ancestors -- the expected time to coalescence is 2Ne generations -- connecting polymorphism data to demographic history. Together these tools turn static genotype data into inferences about selection, migration, and history.

The drift-selection threshold

Selection overpowers drift only when s >> 1/(2Ne). In a population with Ne = 10,000, alleles with selection coefficients below ~0.00005 evolve essentially neutrally -- nearly-neutral theory's central insight, due to Tomoko Ohta.

Key points

  • Hardy-Weinberg (p^2 + 2pq + q^2 = 1) is the null model; its violations reveal evolutionary forces.
  • Selection is quantified by fitnesses; allele frequency change delta-p depends on s, h, and current frequencies.
  • Drift variance is p(1-p)/(2Ne); heterozygosity decays as (1 - 1/(2Ne))^t per generation.
  • Mutation-selection balance maintains deleterious recessives at q-hat near sqrt(mu/s).
  • Migration homogenizes populations; local adaptation requires s to exceed m.
  • Fst measures differentiation; linkage disequilibrium enables association mapping; coalescence dates ancestry.

Common misconceptions

✗ Myth: Hardy-Weinberg describes real populations.

✓ Truth: It describes an idealized null. Real populations violate its assumptions constantly -- that is the point. The equilibrium is a baseline for detecting selection, inbreeding, subdivision, and genotyping error, not a description of nature.

✗ Myth: Selection always wins over drift.

✓ Truth: Only when s >> 1/(2Ne). In small populations drift dominates, and even beneficial mutations are usually lost when rare (fixation probability ~2s). Most molecular evolution may be nearly neutral rather than adaptive.

✗ Myth: A high Fst value proves local adaptation.

✓ Truth: Demographic history -- bottlenecks, founder effects -- also inflates Fst. Genome-wide Fst outliers are candidates for selection, but they require follow-up: functional evidence or replicated environmental correlations.

Frequently asked questions

Why do rare recessive beneficial alleles spread so slowly?

Because they are almost always in heterozygotes where the dominant wild-type allele masks them. Selection can only act efficiently once the allele is common enough for homozygotes to appear, so the initial phase is dominated by drift and can take thousands of generations.

What is effective population size and why is it smaller than census size?

Ne is the size of an idealized population that would drift at the observed rate. Unequal sex ratios, high variance in offspring number, and bottlenecks all reduce Ne below census size -- sometimes by orders of magnitude, as in highly fecund marine species.

How does mutation-selection balance explain persistent genetic diseases?

Deleterious recessive alleles hide in healthy heterozygotes, so selection removes them only slowly while mutation recreates them. The equilibrium frequency q-hat = sqrt(mu/s) is small but nonzero, exactly as observed for diseases like cystic fibrosis and Tay-Sachs.

What does Fst = 0.12 between populations actually mean?

That about 12 percent of total genetic variation is distributed between populations and 88 percent within them. It indicates modest differentiation -- consistent with mostly shared variation plus some local differences -- not deep division.