Heterozygosity

Heterozygosity is the primary quantitative measure of standing genetic variation within a diploid breeding population, representing either the proportion of diploid individuals that possess two distinct alleles at a homologous locus (observed heterozygosity, H_o) or the probability that two alleles drawn at random from the gene pool are non-identical (expected heterozygosity or Nei's gene diversity, H_e). Within the mathematical framework of population genetics, heterozygosity links single-locus Mendelian segregation to genome-wide patterns of genetic drift, mutation, natural selection, and non-random mating.

At the sequence level, expected heterozygosity generalizes to nucleotide diversity (π), which quantifies the average number of pairwise nucleotide differences per site across sampled DNA sequences. Comparing observed heterozygosity against Hardy-Weinberg principle expectations allows evolutionary biologists and conservation geneticists to partition inbreeding within subpopulations, detect cryptic geographic subdivision via the Wahlund effect, and estimate the long-term effective population size (N_e) of natural species.

Interactive Simulation: Explore how finite population size, mutation, and migration alter observed and expected heterozygosity across generations in the Allele Drift Simulator.

Observed Heterozygosity (H_o) and Nei's Gene Diversity (H_e)

In a sample of N diploid individuals genotyped at a single polymorphic locus with k distinct alleles, observed heterozygosity (H_o) is the empirical fraction of individuals carrying two different allelic states (A_iA_j where i ≠ j):

H_o = (1 / N) · ∑_i≠j N_ij = 1 − ∑_i=1^k P_ii

where N_ij is the count of heterozygous A_iA_j individuals and P_ii is the observed frequency of homozygotes for allele A_i. While H_o directly measures the genotypic state of the sampled cohort, it fluctuates strongly under departures from random panmixia such as consanguineous mating, self-fertilization, or assortative mating.

To obtain a metric of allelic diversity that depends only on allele frequencies rather than mating system, Masatoshi Nei (1973) formalized expected heterozygosity (H_e, also termed gene diversity) as the probability that two alleles sampled independently with replacement from the gene pool are different. If allele A_i occurs at frequency p_i (with ∑ p_i = 1), the probability of drawing two identical copies of A_i is p_i², yielding:

H_e = 1 − ∑_i=1^k p_i²

For a biallelic locus with allele frequencies p and q = 1 − p, expected heterozygosity reduces to the familiar Hardy-Weinberg heterozyote proportion H_e = 1 − (p² + q²) = 2pq, which attains a maximum value of H_e = 0.50 when p = q = 0.50. For a multiallelic locus with k alleles at equal frequencies (p_i = 1/k), maximum gene diversity approaches H_e,max = 1 − (1/k). In finite sample sizes of n = 2N chromosomes, an unbiased estimator corrects for downward sampling bias: Ĥ_e = [n / (n − 1)] · (1 − ∑ p̂_i²).

Nucleotide Diversity (π) and Watterson's Estimator (θ_W)

In whole-genome sequencing studies where variation is analyzed per base pair rather than per multiallelic marker, heterozygosity is expressed as nucleotide diversity (π), introduced by Masatoshi Nei and Wen-Hsiung Li (1979). Nucleotide diversity measures the mean number of nucleotide differences per site between any two DNA sequences chosen at random from the population:

π = ∑_i<j 2 · x_i · x_j · π_ij

where x_i and x_j are the frequencies of haplotypes i and j, and π_ij is the proportion of nucleotide sites at which the two sequences differ. Equivalently, if h_l = [ n / (n − 1) ] · (1 − ∑ p_li²) is the site-specific expected heterozygosity at nucleotide position l across a sequence of length L, then π = (1 / L) ∑_l=1^L h_l. Across eukaryotes, π spans more than two orders of magnitude, from π ≈ 0.0001 (0.01%) in bottlenecked island vertebrates to π ≈ 0.001 (0.1%) in Homo sapiens, π ≈ 0.01 (1%) in Drosophila melanogaster, and π > 0.04 (4%) in hyper-diverse marine broadcast spawners such as the sea squirt Ciona savignyi.

Wright's Fixation Indices (F_IS, F_ST, F_IT)

When a diploid population mates at random in the absence of selection, migration, or drift, observed heterozygosity equals expected heterozygosity (H_o = H_e). Sewall Wright (1921, 1951) quantified fractional reductions in observed heterozygosity relative to Hardy-Weinberg expectations using the inbreeding coefficient F:

F = (H_e − H_o) / H_e = 1 − (H_o / H_e)

Rewriting genotype frequencies at a biallelic locus in terms of F yields P(A₁A₁) = p²(1 − F) + pF, P(A₁A₂) = 2pq(1 − F), and P(A₂A₂) = q²(1 − F) + qF. In hierarchical populations composed of multiple local demes (subpopulations), Wright partitioned total heterozygosity deficit into three F-statistics comparing individuals (I), subpopulations (S), and the total metapopulation (T):

1 − F_IT = (1 − F_IS) · (1 − F_ST)

Here, F_IS = (H̄_S − H̄_I) / H̄_S measures the deficiency of observed heterozygotes (H̄_I = H_o) within local subpopulations relative to subpopulation Hardy-Weinberg expectation (H̄_S), reflecting non-random mating such as consanguineous pairing or selfing. Conversely, F_ST = (H_T − H̄_S) / H_T measures the reduction in expected heterozygosity of subpopulations relative to the pooled total metapopulation (H_T) caused by allele frequency divergence across demes.

The Wahlund Effect in Subdivided Populations

Even when every local subpopulation mates at random internally (F_IS = 0), pooling samples across two or more genetically differentiated subpopulations produces an apparent deficit of observed heterozygotes (H_o < H_T) known as the Wahlund effect, formulated by Swedish geneticist Sten Wahlund in 1928. Consider two equally sized subpopulations (k = 1, 2) that are each in internal Hardy-Weinberg equilibrium with allele frequencies p₁ and p₂, and mean allele frequency p̄ = (p₁ + p₂)/2. The average observed heterozygosity across the two demes is:

H̄_o = H̄_S = (1/2) · (2p₁q₁ + 2p₂q₂)

However, if an investigator pools both demes without recognizing the geographic boundary, the expected heterozygosity of the pooled sample is H_T = 2p̄q̄. Expanding the algebraic difference reveals that H̄_o is always smaller than H_T whenever p₁ ≠ p₂:

H_T − H̄_S = 2 · σ_p²

where σ_p² = (1/2)[(p₁ − p̄)² + (p₂ − p̄)²] = (p₁ − p₂)² / 4 is the variance in allele frequency between subpopulations. Dividing this heterozygosity deficit by H_T yields Wright's fixation index F_ST = σ_p² / (p̄q̄). Thus, population subdivision always converts a fraction F_ST of total expected heterozygosity into excess homozygosity across the pooled metapopulation.

Geometric Decay of Heterozygosity in Finite Populations

In a closed, finite Wright-Fisher population of diploid effective size N_e without new mutations, random gametic sampling erodes heterozygosity every generation. When two gametes are drawn from the parental pool to form a zygote in generation t, there is a probability of 1 / (2N_e) that both gametes replicate the exact same parental gene copy (autozygosity by descent), and a complementary probability of 1 − 1 / (2N_e) that they are drawn from two distinct parental gene copies. If they come from distinct copies, their probability of being heterozygous is the heterozygosity of the previous generation, H_t−1:

H_t = [ 1 − 1 / (2N_e) ] · H_t−1

Iterating this recurrence relation from an initial heterozygosity H₀ across t generations gives the exact geometric decay law and its continuous exponential approximation (exp(...)):

H_t = H₀ · [ 1 − 1 / (2N_e) ]^t ≈ H₀ · exp[ −t / (2N_e) ]

The half-life of standing heterozygosity (t_1/2, the number of generations required for H_t to drop to 0.5 H₀) is t_1/2 = 2N_e · ln(2) ≈ 1.386 N_e generations. In small captive populations or island endemics where N_e ≤ 50, half of all initial heterozygosity vanishes within ~69 generations, elevating homozygosity for deleterious recessive alleles and precipitating inbreeding depression.

Mutation-Drift Balance under the Infinite-Alleles Model

In natural populations over evolutionary timescales, the loss of heterozygosity by genetic drift is counteracted by the introduction of novel allelic variants via genetic mutation. In 1964, Motoo Kimura and James F. Crow introduced the infinite-alleles model (IAM), in which every mutation at a locus occurs at rate μ per gamete per generation and generates a novel allele not currently present in the population.

For two alleles sampled in generation t to be identical in state (homozygous, 1 − H_t), neither allele can have mutated in the preceding generation—which occurs with probability (1 − μ)² ≈ 1 − 2μ—and they must either trace back to the same parental copy with probability 1 / (2N_e) or trace back to two distinct identical copies with probability [1 − 1 / (2N_e)] · (1 − H_t−1). Setting H_t = H_t−1 = H_eq and neglecting terms of order μ² yields the fundamental mutation-drift equilibrium equation of the neutral theory of molecular evolution:

H_eq = (4N_eμ) / (1 + 4N_eμ) = θ / (1 + θ)

where θ = 4N_eμ is the population-scaled mutation parameter. When θ ≪ 1 (as applies to single nucleotide sites where μ ≈ 10⁻⁸ per base pair), H_eq ≈ θ = 4N_eμ, meaning equilibrium nucleotide diversity (π) is directly proportional to the product of effective population size and mutation rate. Conversely, under the stepwise mutation model (SMM) developed by Tomoko Ohta and Motoo Kimura (1973) for microsatellite tandem repeats where mutations add or subtract a single repeat unit, equilibrium heterozygosity follows H_eq = 1 − (1 + 8N_eμ)^−1/2.

Why can observed heterozygosity (H_o) differ substantially from expected heterozygosity (H_e)?

Expected heterozygosity (H_e = 1 − ∑ p_i²) depends strictly on allele frequencies in the gene pool, whereas observed heterozygosity (H_o) reflects how those alleles are paired into diploid genotypes. When H_o < H_e (F > 0), the population exhibits a heterozygote deficit caused by inbreeding (selfing or consanguineous mating), positive assortative mating, null alleles during PCR genotyping, or unacknowledged geographic population structure (the Wahlund effect). Conversely, when H_o > H_e (F < 0), the population exhibits a heterozygote excess, which can arise from heterozygote advantage (overdominant selection against homozygotes), negative assortative (disassortative) mating, or small Dioecious breeding cohorts where allele frequencies differ by chance between males and females.

How does heterozygosity-fitness correlation (HFC) relate to inbreeding depression?

Across wild animal and plant cohorts, multi-locus individual heterozygosity measured across neutral genome-wide markers frequently correlates positively with survival, growth rate, fecundity, and parasite resistance—a phenomenon termed a heterozygosity-fitness correlation (HFC). Two primary mechanisms generate HFCs: (1) the general effect hypothesis, in which genome-wide marker heterozygosity serves as an inverse proxy for the individual's pedigree inbreeding coefficient (F), reflecting reduced expression of deleterious recessive alleles across thousands of unlinked loci (inbreeding depression); and (2) the local effect hypothesis, in which neutral markers sit in genetic linkage disequilibrium with specific functional loci (such as Major Histocompatibility Complex immune genes) maintained under heterozygote advantage.

Why is human nucleotide diversity (π ≈ 0.001) much lower than that of chimpanzees or fruit flies despite a human census size of 8 billion?

Equilibrium heterozygosity (H_eq ≈ 4N_eμ) reflects the long-term harmonic mean effective population size (N_e) over the past ~4N_e generations rather than the current census population (N). Because Homo sapiens experienced a Pleistocene population bottleneck and serial founder effects during the Out-of-Africa expansion ~50,000–70,000 years ago, human long-term N_e is only ~10,000 breeding individuals. Because new neutral mutations require on average ~4N_e generations to reach mutation-drift equilibrium, the demographic expansion of humans following the Neolithic agricultural revolution (~400 generations ago) is far too recent to have restored equilibrium nucleotide diversity, although it has generated an abundance of very rare singleton variants.

1. Nei, M. (1973). "Analysis of gene diversity in subdivided populations." Proceedings of the National Academy of Sciences, 70(12), 3321–3323. doi:10.1073/pnas.70.12.3321
2. Wright, S. (1951). "The genetical structure of populations." Annals of Eugenics, 15(1), 323–354. doi:10.1111/j.1469-1809.1949.tb02451.x
3. Kimura, M., & Crow, J. F. (1964). "The number of alleles that can be maintained in a finite population." Genetics, 49(4), 725–738. doi:10.1093/genetics/49.4.725
4. Nei, M., & Li, W.-H. (1979). "Mathematical model for studying genetic variation in terms of restriction endonucleases." Proceedings of the National Academy of Sciences, 76(10), 5269–5273. doi:10.1073/pnas.76.10.5269