1 Overview of linked loci and selection
Selection on linked loci describes evolutionary change in which natural or artificial selection acting at one genetic position alters allele frequencies at neighboring positions. This occurs because loci that are physically close on a chromosome are more likely to be inherited together than distant loci. As a result, the genomic pattern of genetic variation reflects not only the selected site itself, but also the extent to which neighboring alleles are “carried” during the selective process or are removed along with selected alleles.
1.1 Linkage, inheritance, and linkage disequilibrium
Linkage refers to the non-random association of alleles at different loci due to physical proximity on a chromosome. During meiosis, recombination can break these associations, but when recombination is limited relative to selection, linked inheritance becomes strong. The statistical consequence is linkage disequilibrium (LD): a deviation from the expectation that alleles at different loci occur together according to their marginal frequencies. LD can be transient, generated by selection, or long-lived when recombination fails to erase associations efficiently.
1.2 Selection types that create linked-locus effects
Linked-locus effects arise when selection changes allele frequencies at a focal locus and those allele-frequency changes propagate to nearby loci. Two broad categories are often distinguished. In “selective sweep” scenarios, a beneficial allele increases rapidly in frequency and nearby haplotypes can rise in frequency with it. In “background selection” scenarios, purifying selection removes deleterious alleles; neutral or weakly selected variants linked to them can also experience reduced diversity without any single advantageous mutation driving the pattern.
1.3 Timescales: short-term selective sweeps vs long-term dynamics
The impact of linked selection depends on time horizon. Selective sweeps are typically analyzed over the period when the favored allele rises to high frequency; during this window, LD can intensify and haplotypes near the target locus can show sharp reductions in variation. Background selection and other forms of purging act continuously, producing more gradual depletion of diversity across functional genomic regions. Over long spans, repeated episodes of selection, demographic change, and recombination collectively shape the final genomic landscape.
2 Mechanisms: how selection at one locus affects others
The central reason nearby loci change is that alleles are packaged into haplotypes. Selection alters the success of haplotypes carrying particular alleles, and recombination determines how readily those haplotypes are reshuffled. Several mechanisms explain how these processes yield correlated allele-frequency changes across linked loci.
2.1 Genetic hitchhiking
Genetic hitchhiking occurs when an allele increases in frequency because it is beneficial, and neighboring alleles “ride along” because they are frequently found on the same chromosome background.
2.1.1 Selective sweep of a beneficial allele
During a sweep, the favored variant and the haplotype carrying it become overrepresented. If recombination is not frequent enough to detach neighboring alleles from the sweeping background, those linked alleles increase in frequency together, regardless of whether they are beneficial, neutral, or deleterious. The result is a localized genomic region with reduced variation and altered haplotype structure.
2.1.2 Effects on neutral and deleterious neighbors
Hitchhiking can increase neutral variation on the sweeping haplotype (by bringing it to higher frequency), but it can also reduce diversity because multiple ancestral haplotypes collapse toward those linked to the successful allele. For deleterious neighbors, hitchhiking can temporarily raise their frequency if they are carried by the favorable haplotype; later, purifying selection may counteract their persistence, especially if recombination has generated new associations that expose them to selection.
2.2 Background selection (purging of linked variation)
Background selection describes how purifying selection against deleterious variants indirectly reduces diversity at nearby loci. Rather than relying on one strongly beneficial mutation, it depends on the continual presence of harmful alleles and the steady removal of haplotypes carrying them.
2.2.1 Removal of deleterious alleles and linked diversity reduction
When deleterious alleles reduce the fitness of their carriers, chromosomes carrying them are less likely to contribute to future generations. Nearby neutral alleles can be lost as collateral because they are frequently associated with the harmful variants. Even if the neutral loci are not themselves affected by selection, their effective representation in the population declines when linked haplotypes are repeatedly eliminated.
2.2.2 Consequences for allele frequency spectra
Because background selection repeatedly removes certain haplotypes, the allele frequency spectrum at linked neutral sites can shift. Typically, the number of low-frequency variants decreases less predictably than high-frequency variants are altered; the overall effect is often a reduction in diversity relative to neutral expectations. The precise shape depends on the distribution of fitness effects, recombination rate, and demography.
2.3 Hill–Robertson interference
Hill–Robertson interference refers to the reduction in the efficacy of selection caused by multiple selected sites competing within a finite population. Linkage creates correlations among sites, so selection at one locus can hinder the success of beneficial alleles at another locus (and similarly complicate purging).
2.3.1 Clonal interference and linkage effects (conceptual overview)
Even without literal asexual reproduction, linkage can create “competition” among different haplotypes carrying beneficial mutations. When multiple advantageous variants occur in the same population, different lineages can expand simultaneously. If recombination is insufficient to combine beneficial alleles into more fit haplotypes, some advantageous backgrounds may be outcompeted before they fix, lowering the overall rate at which adaptation proceeds. This interference is conceptually related to clonal interference in asexual populations.
2.3.2 Multiple selected sites and reduced efficacy of selection
With many selected positions, beneficial changes can be trapped on suboptimal backgrounds, while deleterious alleles on the same haplotype can limit their spread. Conversely, deleterious mutations can persist longer if they repeatedly hitchhike with beneficial ones. The net effect is a genome-wide alteration of the effective selection coefficient relative to a model where sites are independent.
2.4 Recombination-mediated decoupling
Recombination reshuffles genetic material and can separate selected alleles from their neighbors. This “decoupling” determines whether linked loci remain correlated long enough for selection to create noticeable patterns.
2.4.1 Recombination rate and the extent of “drag” along the chromosome
The extent of hitchhiking or diversity reduction is strongly tied to recombination rate. Low recombination produces longer linkage blocks and larger footprints around selected sites, because haplotypes remain intact. High recombination shortens footprints by frequently breaking associations between loci, limiting how far the influence of selection extends.
2.4.2 Threshold-like transitions in linkage regimes
Many analytical treatments yield regime-like behavior: when recombination is much weaker than selection-driven changes, linked effects are strong; when recombination is sufficiently high, loci behave more independently. The transition is not universally sharp across all models, but the qualitative shift is often characterized by a critical balance between recombination and the timescale of selection.
3 Population-genetic models and key quantities
Population-genetic models formalize how selection, linkage, recombination, and drift jointly determine allele and haplotype trajectories. Key quantities summarize the strength of selection, the structure of linkage, and the dynamics of correlations.
3.1 Effective selection coefficients for linked regions
Because selection at a focal locus influences the fate of linked haplotypes, the strength of selection “felt” by neighboring sites is modified relative to the focal site’s own fitness effect. This motivates the concept of effective selection coefficients for linked regions, which account for the probability that recombination decouples a neutral or weakly selected allele from the selected background. Under some approximations, effective selection can be reduced due to interference, or it can be expressed through an “effective population size” or “effective fitness” for linked segments.
3.2 Fitness landscapes with multiple loci
Multi-locus models represent fitness as a function of alleles across positions. Additive fitness effects, dominance, epistasis, and gene–gene interactions shape which haplotypes are favored. Linkage then influences whether beneficial combinations arise and persist. Under epistasis, recombination can either help by combining compatible alleles or hinder by breaking beneficial combinations.
3.3 Deterministic vs stochastic descriptions
Deterministic models track expected changes in allele frequencies under infinite-population assumptions. Stochastic descriptions incorporate random drift and demographic fluctuations, which can be crucial when beneficial alleles are rare, when selected haplotypes are limited in number, or when recombination is too infrequent to prevent loss by chance.
3.3.1 Drift in finite populations and its interaction with linkage
In finite populations, drift can eliminate beneficial lineages before they reach high frequency, and it can also preserve deleterious haplotypes longer than deterministic selection would predict. Linkage can amplify these effects because drift acting on one haplotype influences multiple loci simultaneously. The combined influence can lead to variable sweep outcomes and inconsistent signatures between replicate populations.
3.4 Linkage disequilibrium dynamics under selection
Selection generates LD by making certain allele combinations disproportionately successful. Recombination reduces LD by breaking up these combinations. Many models track the time evolution of LD around selected sites, including how LD phase (positive or negative association) and magnitude depend on selection strength and recombination.
3.4.1 Phase and magnitude of LD changes near selected sites
The direction of LD depends on whether the selected allele tends to occur on background haplotypes carrying correlated variants. In sweep-like conditions, LD often increases during the rise of the favored haplotype and decays afterward as recombination breaks associations. The spatial profile of LD across distance from the selected locus reflects the competing processes of haplotype growth and recombination erosion.
3.5 Parameters used to summarize genomic outcomes
Empirical and theoretical work often reduces complex dynamics to a set of summary statistics describing haplotypes, diversity, and correlation.
3.5.1 Haplotype structure and summary statistics
Haplotype-based quantities include haplotype length, the number of distinct haplotypes within a region, and measures of haplotype homozygosity. These statistics capture patterns such as long, high-frequency haplotypes near sweep sites or reduced haplotype diversity following purging events.
3.5.2 Diversity (e.g., heterozygosity) and divergence patterns
Diversity metrics include heterozygosity, nucleotide diversity, and measures based on coalescence time. Divergence between populations can also reflect linked selection when selection changes the ancestral composition of genomic segments. Differences in diversity between selected and neutral regions often provide clues about the underlying mechanism.
4 Genomic signatures of selection on linked loci
Because linked selection changes haplotype structure and coalescent histories, it leaves characteristic signatures in genomic data. The signatures vary between hitchhiking and background selection and can be modulated by recombination and the density of functional sites.
4.1 Patterns of reduced genetic diversity
Selection reduces variation by altering coalescent times and eliminating certain lineages more frequently than others.
4.1.1 Selective sweep signatures around the target locus
In a selective sweep, a region surrounding the advantageous allele often shows reduced diversity, an excess of rare variants, and a pattern of extended haplotypes. The reduction can be asymmetric if selection occurred on a standing haplotype or if demographic events affected allele histories. The hallmark is a localized deficit of polymorphism accompanied by distinctive changes in haplotype lengths.
4.1.2 Background selection signatures along regions with functional density
Background selection tends to lower diversity over broader spans, especially in genomic segments enriched for functional elements. Rather than showing one sharply defined boundary around a single beneficial locus, the pattern often reflects a gradient: stronger diversity depletion in regions with higher density of deleterious targets and reduced depletion where functional content is sparse.
4.2 Changes in allele frequencies and haplotype length
Selection at one locus reshapes the distribution of allele frequencies and the extent of haplotypes that carry the selected background.
4.2.1 Extended haplotypes and “sweep islands”
Sweep islands refer to genomic neighborhoods where haplotypes are unusually long and uniform due to a recent selective event. These regions can be detected by comparing haplotype-based statistics to neutral expectations that incorporate recombination and demographic history.
4.2.2 Skewed site-frequency distributions
Selective sweeps can skew site-frequency distributions, often increasing the proportion of low-frequency variants near the selected site because older lineages have been eliminated while new mutations since the sweep remain mostly rare. The exact spectrum depends on sweep strength, timing, and the influence of recombination.
4.3 Correlated variation across loci
Linked selection produces predictable correlations between loci and between functional and neutral sites.
4.3.1 LD decay patterns near selected regions
Around a selected locus, LD often decays more slowly or exhibits a structured profile compared with genome-wide averages. As recombination breaks associations, the LD signal diminishes with distance, producing a characteristic decay curve that can reveal the strength and timing of selection.
4.3.2 Covariation among functional and neutral sites
In background selection, functional sites under purifying selection can correlate with reduced diversity at nearby neutral loci. If many functional elements contribute to purging, then covariation across multiple sites and pathways becomes visible in genome scans that jointly analyze functional and nonfunctional regions.
4.4 Comparison among selection models
Different linked-selection processes can produce overlapping signals, so comparison requires attention to expected patterns across distance, frequency spectra, and recombination dependence.
4.4.1 Distinguishing hitchhiking vs background selection
Hitchhiking is typically localized around a specific allele whose increase drives a sweep-like reduction in variation. Background selection often yields broader, more continuous reductions tied to functional density and purifying selection intensity. The contrast can be assessed by whether diversity is depleted in a narrow region versus a wider neighborhood and by whether haplotype structure shows sweep-like “islands” or smoother gradients.
4.4.2 Recognizing recombination-limited footprints
If recombination is low, the footprints of selection extend farther. Therefore, observing how the magnitude of diversity reduction changes with recombination rate can help distinguish linked-selection outcomes from purely demographic causes or unlinked selection.
5 Special cases and extensions
Real genomes and traits can depart from simplified assumptions of independence, single-locus selection, or simple fitness effects. Extensions address standing variation, dominance and epistasis, multi-locus selection, and population structure.
5.1 Selection on standing genetic variation
Selection can act on alleles already present in the population rather than on new mutations. This alters timing and the shape of linked signatures.
5.1.1 Linked changes when beneficial alleles pre-exist
Standing variants can rise to high frequency without the long waiting time typical of de novo mutation sweeps. The resulting LD and diversity patterns may be less extreme in some regions because the allele’s age and pre-existing haplotype diversity already reflect older recombination. Linked changes can still occur, but the signature may be smoother or shifted in time.
5.1.2 Limits imposed by recombination
Even with standing variation, recombination determines how far the selected allele’s rise drags linked neighborhoods. If the allele originated on multiple haplotypes, selection can increase several backgrounds, reducing the expected degree of homogenization. Conversely, if it was concentrated on a narrow haplotype, sweep-like effects remain strong.
5.2 Partial dominance and epistasis across linked loci
Dominance and epistasis influence how selection acts on haplotypes and therefore how linked loci respond. Dominance can affect whether heterozygotes are favored or disfavored; epistasis can make fitness depend on allele combinations.
5.2.1 Dominance effects on haplotype spread
If a beneficial allele is partially recessive, it may increase more slowly when rare because heterozygotes gain less fitness. This slower spread can alter the timing of LD buildup and the extent of haplotype homogenization. Dominance can also influence whether selection maintains multiple haplotypes for longer periods.
5.2.2 Epistatic interactions and multi-locus selection
Epistasis can couple loci so that fitness depends on combinations of alleles. In such cases, recombination may either generate more favorable combinations (accelerating adaptation) or disrupt them (reducing the success of lineages). Linked selection signatures therefore depend not only on physical proximity but also on the structure of functional interactions.
5.3 Multiple loci under simultaneous selection
When selection acts at many positions, linked selection becomes a collective outcome rather than a series of isolated episodes.
5.3.1 Layered sweeps and overlapping selective footprints
If multiple beneficial alleles sweep in close temporal proximity, their footprints can overlap, producing complex haplotype and diversity patterns. Overlapping sweeps can mask each other or create intermediate signals. Background selection may also interact with sweeps, complicating inference about which mechanism dominates.
5.3.2 How polygenic traits shape linkage patterns
For polygenic traits influenced by many loci of small effect, selection may be spread across the genome. This can generate broad shifts in allele frequencies without pronounced sweep footprints. The resulting LD changes may be subtle but detectable through patterns in cumulative haplotype structure or correlated frequency shifts across many sites.
5.4 Population structure and migration
Population history affects linked selection signals by altering allele mixing and the effective trajectories of selected haplotypes.
5.4.1 How mixing of subpopulations affects LD under selection
Migration and admixture introduce new haplotypes and can reduce or reconfigure LD by mixing ancestry. Selection acting within structured populations can create different linked footprints in different subgroups, and subsequent mixing can blur these patterns in aggregate samples.
5.4.2 Local vs global selection signatures
A selection event can produce strong local signatures within a subpopulation while appearing weaker when all populations are pooled. This distinction motivates analyses that allow spatial or demographic heterogeneity when interpreting linked-locus signatures.
6 Empirical approaches and inference (high level)
Empirical inference uses patterns of haplotypes, LD, diversity, and recombination to detect and discriminate linked selection. Because many processes can generate similar signatures, robust methods incorporate recombination and demography.
6.1 Designing tests using linkage information
Tests can explicitly leverage linkage structure by using recombination maps, inferred ancestral recombination rates, or models that predict how selection should decay with distance.
6.1.1 Incorporating recombination maps
Recombination maps translate physical distance into expected recombination rates, enabling models to predict the size and shape of linked-selection footprints. Accurate maps improve the calibration of distance-dependent expectations, reducing false positives driven by incorrect linkage assumptions.
6.2 Using haplotype-based vs site-based summary methods
Haplotype-based methods often provide greater sensitivity to recent or strong linked sweeps because they capture phase information and the persistence of extended haplotypes.
6.2.1 When haplotypes provide stronger signals than single-site data
Single-site statistics can be ambiguous because allele-frequency changes may arise from demography or unlinked selection. Haplotype statistics can distinguish scenarios by identifying coordinated shifts across multiple linked sites and by detecting long-range LD structures consistent with hitchhiking.
6.3 Model-based inference and simulation
Model-based inference compares observed patterns to expectations generated under linked selection, with key uncertainties addressed through simulation or approximate methods.
6.3.1 Forward-time simulation for linked selection
Forward-time simulations track haplotype evolution under explicit linkage, recombination, selection, and demography. They are particularly useful for exploring complex multi-locus scenarios, though they can be computationally demanding.
6.3.2 Approximate methods for fast genome scans
Approximate approaches aim to accelerate genome scanning by using summary statistics, likelihood approximations, or coalescent-based shortcuts. These methods trade some mechanistic realism for speed, making them suitable for large datasets but requiring careful calibration.
7 Practical considerations and limitations
Inference depends on data quality, demographic assumptions, recombination-rate accuracy, and the interpretation of signals in dense functional regions.
7.1 Effects of demographic history (conceptual disentangling)
Population size changes, bottlenecks, expansions, and structure can mimic selection signatures by altering coalescent times and allele-frequency spectra. A key practical task is conceptual disentangling: determining whether linked-locus patterns reflect selection-induced haplotype structure or demographic processes that produce similar reductions in diversity.
7.2 Recombination map uncertainty
If recombination rates are misestimated, the predicted extent of linked footprints can be wrong, weakening power or inflating error rates. Uncertainty is especially problematic in regions with sparse recombination-rate observations or in species with limited recombination map coverage.
7.3 Sampling strategy and statistical power
Power depends on sample size, geographic representation, sequencing depth, and the accuracy of haplotype phasing. Under-sampling can obscure LD and haplotype structure; over-sampling without appropriate model adjustment can introduce biases if related individuals are included.
7.4 Interpreting signals in regions with many functional elements
Genomic regions with dense functional content often show patterns consistent with background selection even without any single recent sweep. Distinguishing between frequent purging and occasional hitchhiking requires attention to how signals distribute across distance, recombination rate, and functional categories.
8 Glossary of core terms
8.1 Linkage, LD, haplotype, and selective sweep
Linkage: Physical proximity of loci on a chromosome that increases their likelihood of being inherited together. Linkage disequilibrium: Non-random association between alleles at different loci. Haplotype: A combination of alleles along a chromosome segment inherited as a unit. Selective sweep: Rapid increase in frequency of a beneficial allele, often reducing diversity and reshaping haplotypes nearby.
8.2 Background selection and hitchhiking
Background selection: Purifying selection against deleterious variants that indirectly reduces variation at linked neutral sites. Hitchhiking: Indirect change of allele frequencies at loci linked to a site under positive selection, due to the shared inheritance of haplotypes.
8.3 Recombination rate and linkage blocks
Recombination rate: The probability per unit time or distance that recombination separates genetic loci, controlling how quickly LD breaks down. Linkage block: A genomic segment where loci are strongly correlated because recombination between them is relatively rare.