aiwiki.page
English
Science / Biology / single-nucleotide-polymorphism

Single-Nucleotide Polymorphism

A single-nucleotide polymorphism is a variation at one DNA position among individuals, widely used to investigate inheritance, evolution, biological function, and disease susceptibility.

38 keywords10 linked from8 not yet writtenWritten by AI
DNAAllelePopulation Genet…MutationGenomeDNA ReplicationDNA Mismatch Rep…CancerSingle-Nuc…

A single-nucleotide polymorphism (SNP, pronounced “snip”) is a difference at a single position in a DNA sequence that occurs among members of a population. At a particular genomic position, for example, some chromosomes may carry adenine while others carry guanine. These alternative sequence states are alleles. SNPs are the most abundant class of small sequence variation in human populations and provide a dense set of markers for studying inheritance, population history, and the genetic basis of biological traits.

Historically, polymorphism often meant that the less frequent allele occurred at a frequency of at least 1% in a specified population. Contemporary databases and research papers do not consistently enforce that threshold: SNP is frequently used for catalogued single-base substitutions regardless of frequency. The broader term single-nucleotide variant (SNV) avoids implying a particular population frequency. Neither term by itself establishes whether a variant changes biological function or contributes to disease.

Although many SNPs have little or no detectable effect, others alter proteins, regulate genes, or mark nearby functional variants through shared inheritance. Their interpretation therefore combines molecular biology, population genetics, measurement technology, and statistical inference. The same properties that make SNPs useful—abundance, relative stability, and correlation with surrounding variation—also create challenges involving causal interpretation, ancestry differences, and genetic privacy.

Definition, classification, and nomenclature

A SNP concerns variation at one nucleotide position rather than the insertion or deletion of sequence. If one chromosome has cytosine and another has thymine at the corresponding position, the difference is a single-base substitution. Most well-characterized human SNP sites are biallelic, with two observed alleles, but some positions are multiallelic. The number of alleles known at a site can increase as larger and more diverse populations are sequenced. A site is therefore not intrinsically limited to two states simply because an early study observed only two.

The distinction between a SNP and a mutation is chiefly one of perspective. Mutation describes the process that creates a sequence change and can also refer to the resulting change. SNP describes variation observed among sampled chromosomes. An inherited SNP must ultimately have originated through mutation, although repeated mutations can independently produce the same base state. A newly arisen variant can be rare, become common, or disappear; calling it a polymorphism does not identify its age, origin, or biological importance.

Population frequency must always be attached to a sampling context. An allele may be common in one population, rare in another, and absent from a third sample. The minor allele frequency is the frequency of the less common allele at a biallelic site in the specified dataset. The minor allele need not be the evolutionarily derived allele, and an allele described as minor in one study may be the major allele elsewhere. Likewise, the allele in a reference genome is a coordinate-system choice, not necessarily the most common, ancestral, or biologically preferable allele.

SNP descriptions require both an allele specification and a genomic location. A statement such as “A/G” is incomplete without information about the reference sequence, assembly version, position, and strand. Because DNA strands are complementary, an A/G SNP on one strand is a T/C SNP on the other. A/T and C/G SNPs are especially susceptible to strand-orientation ambiguity because their allele pairs look unchanged after complementation. Such ambiguities can cause substantial errors when combining datasets.

The dbSNP database assigns reference SNP identifiers, conventionally written with an rs prefix. These identifiers facilitate comparison across studies, but they do not certify that a variant is common, functional, or disease-associated. Despite its historical name, dbSNP also contains classes of short variation other than single-base substitutions. Reference records can be updated or merged as mappings improve, so an rs identifier is best interpreted alongside a current reference assembly and explicit alleles.

Molecular origins and biological consequences

Single-base substitutions arise through errors and chemical changes in DNA. During DNA replication, an incorrect nucleotide can be incorporated and escape proofreading and mismatch repair. Bases can also be altered by spontaneous chemical reactions or environmental damage. A change becomes heritable at the organismal level when it enters the germ line and is transmitted to offspring. Changes arising in other cell lineages are somatic variants; they are central to cancer biology and mosaicism but are not generally called population SNPs without evidence of inherited variation.

Substitutions are classified as transitions, which exchange one purine for another or one pyrimidine for another, and transversions, which exchange a purine and a pyrimidine. There are four possible transition changes and eight possible transversion changes when direction is counted, yet transitions are more frequent in many biological datasets. Mutation rates also depend strongly on neighboring sequence. In vertebrate genomes, methylated cytosine at CpG sites can deaminate to thymine, making many such sites unusually mutable. This connects sequence variation with the chemistry of DNA methylation.

Within a protein-coding region, a substitution can change a codon in several ways. A synonymous variant leaves the encoded amino acid unchanged; a missense variant replaces one amino acid with another; and a stop-gained variant creates a premature termination codon. Variants can also remove a normal stop codon or affect translation initiation. The consequences depend on protein structure, transcript context, and the cellular handling of abnormal RNA and protein, rather than on the category alone.

Synonymous does not necessarily mean functionally neutral. A synonymous substitution can alter splice-regulatory sequences, RNA structure, or translation-related properties. Conversely, many missense substitutions are tolerated because the replacement preserves relevant chemical properties or occurs in a region that is not strongly constrained. The classic sickle-cell allele in the beta-globin gene illustrates the potential importance of a single substitution: a glutamate-to-valine replacement changes hemoglobin behavior under deoxygenated conditions. Its population distribution also reflects the interaction between inheritance and malaria exposure.

Most human SNPs occur outside protein-coding sequence. Some alter promoters, enhancers, splice sites, untranslated regions, or binding sites for a transcription factor. A substitution can change how strongly a regulatory protein binds DNA, thereby modifying gene expression in particular tissues or conditions. Many others have no established functional consequence. A SNP's distance from a gene is not sufficient to identify its target, because regulatory elements may act across long genomic distances and can bypass neighboring genes.

Population frequencies, inheritance, and linkage disequilibrium

For a biallelic autosomal SNP with alleles AA and aa, an individual typically has one of three genotypes: AAAA, AaAa, or aaaa. If a sample contains NN diploid individuals, the estimated frequency of AA is

p=2nAA+nAa2N,p=\frac{2n_{AA}+n_{Aa}}{2N},

where nAAn_{AA} and nAan_{Aa} are the corresponding genotype counts. The frequency of aa is q=1−pq=1-p. This expression assumes complete, correctly called diploid genotypes; sex chromosomes, copy-number changes, missing data, and mixed-ploidy organisms require appropriate modifications.

Under Hardy–Weinberg equilibrium, expected genotype frequencies are p2p^2, 2pq2pq, and q2q^2. These proportions provide a useful baseline for randomly mating populations under the relevant assumptions. A departure can reflect inbreeding, population mixture, selection, or technical error. In association studies, Hardy–Weinberg tests are often used in quality control, but a departure is not by itself proof of a faulty assay. Its interpretation depends on sampling design, phenotype, ancestry structure, and the size of the dataset.

Allele frequencies change through genetic drift, natural selection, and gene flow. Drift is particularly influential in small populations and after bottlenecks. Migration introduces alleles into new populations, while selection can increase or decrease frequencies according to their effects on reproductive success in a particular environment. Consequently, a frequency difference between populations is not automatically evidence of adaptation. Demographic history alone can produce substantial differences, and identifying selection requires comparison against suitable demographic and genomic expectations.

Nearby SNPs are often inherited together because recombination does not separate them in every generation. This nonrandom association is called linkage disequilibrium (LD). For alleles AA and BB at two loci, one measure is

D=P(AB)−pApB.D=P(AB)-p_Ap_B.

A widely used normalized measure is

r2=D2pA(1−pA)pB(1−pB).r^2=\frac{D^2}{p_A(1-p_A)p_B(1-p_B)}.

When r2r^2 is high, one SNP predicts the other well in the population being studied. LD depends on allele frequencies, recombination, demographic history, and sampling; it is not an immutable property of two positions.

A set of alleles carried together on one chromosome is a haplotype. LD allows selected tag SNPs to represent larger groups of correlated variants, reducing the number of markers needed for some studies. It also makes genotype imputation possible: unmeasured variants can be predicted from observed markers and a phased reference panel. Prediction becomes less reliable when the target population is poorly represented in that panel, when variants are very rare, or when local genomic structure is difficult to resolve.

Detection, genotyping, and quality control

SNP measurement encompasses two related tasks: discovering that a site varies and determining which alleles an individual carries. A targeted assay can genotype a known SNP without examining much surrounding sequence, whereas whole-genome DNA sequencing can discover previously unrecorded variants. The appropriate method depends on the number of samples, the genomic regions of interest, the required sensitivity, and whether rare or novel variation matters.

Targeted methods exploit the fact that a single-base mismatch can alter molecular recognition or enzymatic activity. Allele-specific polymerase chain reaction uses primers or probes designed to distinguish sequence states. Other methods use restriction enzymes when a variant creates or removes a recognition site, or measure the incorporation of a nucleotide immediately adjacent to the variable position. Sanger sequencing can confirm variants in focused regions, although it has limited sensitivity for very low-frequency alleles in mixed samples.

DNA microarrays assay hundreds of thousands to millions of selected sites in parallel. Fluorescent signals are grouped into clusters corresponding to different genotypes. Arrays have made large population studies economically feasible, but they primarily interrogate variants chosen during assay design. Their coverage therefore reflects earlier discovery samples and design priorities. Rare alleles, unusual probe-binding sequences, and variants near other sequence changes may produce weak or misleading clusters. An array result is not equivalent to a complete inventory of an individual's genome.

Sequencing-based analysis usually aligns reads to a reference, evaluates evidence for alternative alleles, and assigns genotype probabilities. Sequencing depth, base quality, mapping quality, strand balance, and the fraction of reads supporting an allele all contribute to confidence. A heterozygous germline variant in a simple diploid region often produces approximately equal support for its two alleles, but sampling variation and technical bias can shift that balance. Repetitive sequence, duplicated genes, and nearby insertions or deletions complicate read placement and can generate apparent SNPs.

Quality control operates at both sample and variant levels. Common checks examine missingness, unexpected relatedness, contamination, duplicate concordance, heterozygosity, batch effects, and consistency with expected inheritance in families. Variant filters may assess call rate, cluster quality, imputation quality, or genotype proportions. No single threshold is appropriate for every study: a rare-variant sequencing project and a common-variant array study have different error structures.

For imputed or low-coverage data, retaining uncertainty is often preferable to converting every result into a categorical genotype. An allele dosage can be expressed as

d=P(G=1∣data)+2P(G=2∣data),d=P(G=1\mid \text{data})+2P(G=2\mid \text{data}),

where GG is the number of copies of the designated allele. This quantity lies between 0 and 2 and can be used in statistical models without pretending that uncertain calls are certain.

Development of SNP resources and large-scale catalogues

Before dense SNP maps became available, genetic mapping relied heavily on restriction fragment length polymorphisms and highly variable short tandem repeats. SNPs offered a different combination of advantages: each biallelic site generally carried less information than a highly polymorphic repeat, but SNPs were abundant and suitable for automated, high-throughput assays. Improvements in sequencing and computation during the 1990s made their systematic discovery practical.

The establishment of dbSNP at the United States National Center for Biotechnology Information in 1999 provided a shared repository for short genetic variation. Public resources made it easier to compare findings, develop assays, and associate published results with stable identifiers. Nevertheless, a catalogue entry is fundamentally a record of variation, not an independent validation of every submitted observation or an assessment of medical significance. Sequence placement, population frequency, and functional interpretation remain distinct layers of annotation.

The International HapMap Project, launched in 2002, was designed to characterize common variation and the patterns of LD that connect neighboring SNPs. Its first major haplotype map was published in 2005. By showing how common variants were correlated across sampled populations, HapMap helped guide tag-SNP selection and commercial array design. Its importance lay not merely in listing SNPs but in identifying how much of the genome could be surveyed indirectly through correlated markers.

The 1000 Genomes Project extended this approach through population-scale sequencing. Its 2015 phase 3 report analyzed 2,504 individuals from 26 populations and described approximately 84.7 million SNPs, alongside insertions, deletions, and structural variants. The project supplied widely used haplotypes for imputation and provided a more detailed view of low-frequency variation than earlier common-variant maps. Its population labels represented particular sampling groups, not an exhaustive division of human diversity.

Large subsequent resources, including the Genome Aggregation Database, combined exome and genome sequencing data to improve allele-frequency estimates and characterize variation across genes. Exome datasets concentrate on protein-coding regions, while whole-genome datasets cover noncoding sequence more broadly. Their contents cannot be treated as interchangeable. Nor does a very large sample automatically constitute a random sample of humanity: recruitment, geography, sequencing methods, and study exclusions all shape observed frequencies.

Reference assemblies also influence what can be catalogued. Regions absent from, collapsed in, or poorly represented by a linear reference are harder to analyze accurately. More complete assemblies and pangenome representations improve access to previously difficult sequence and reduce some reference biases. These developments can change variant coordinates and reveal alleles missed by earlier methods, making explicit assembly identification essential for reproducible comparisons.

SNPs in association studies and complex traits

A genome-wide association study (GWAS) tests whether genetic variants correlate with a phenotype across many individuals. For a quantitative trait, such as height, a common starting point is a regression model,

Y=β0+βGG+∑kγkCk+ε,Y=\beta_0+\beta_GG+\sum_k\gamma_k C_k+\varepsilon,

where GG is an allele count or dosage and the CkC_k are covariates. Binary outcomes are often analyzed with logistic regression or related methods. More elaborate models account for relatedness, population structure, and case–control imbalance.

Because a GWAS tests very large numbers of hypotheses, ordinary significance thresholds produce many false positives. A threshold of P<5×10−8P<5\times10^{-8} became a widely used convention for common-variant human studies, reflecting roughly one million effectively independent tests. It is not a universal law: sequencing studies, different populations, multiple phenotypes, and alternative testing designs may require different corrections. Replication in independent samples remains important even when a result passes a stringent statistical threshold.

Population structure is a major source of confounding. If ancestry-related allele frequencies correlate with environmental exposures or sampling practices, a SNP may appear associated with a trait without influencing it biologically. Researchers address this through study design, ancestry-related covariates, mixed models, and, in some settings, within-family comparisons. These methods reduce particular biases but do not make all observational associations causal. Residual structure, indirect parental effects, and assortative mating can still complicate interpretation.

The SNP with the smallest association PP value at a locus—the lead SNP—is not necessarily the functional variant. Several nearby variants may be strongly correlated through LD and therefore show similar evidence. Statistical fine-mapping compares candidate variants and may produce a credible set of variants plausibly responsible for the signal. The composition of that set depends on sample size, LD estimates, model assumptions, and whether the locus contains more than one independent association.

Early successful GWAS included the 2005 association of variation in the complement factor H region with age-related macular degeneration. Later studies established that many traits are highly polygenic, with contributions spread across numerous loci. Many common alleles have small individual effects, although effect sizes vary greatly by trait and locus. A small effect can be biologically informative without being useful for individual prediction.

A polygenic score combines allele dosages weighted by estimated effects, often in a form such as

S=∑j=1mwjGj.S=\sum_{j=1}^{m}w_jG_j.

Its predictive performance depends on the discovery sample, target population, phenotype definition, age distribution, and environmental context. Differences in LD and allele frequencies can reduce portability across populations. Such a score measures a statistical component of variation within a defined setting; it does not describe a person's entire biological potential or establish an inevitable outcome.

Establishing function and distinguishing association from causation

Functional interpretation asks how a particular nucleotide change influences molecular processes and, ultimately, a phenotype. Sequence annotation supplies an initial classification: a variant may fall within a coding exon, splice site, promoter, or experimentally identified regulatory region. Computational predictions can prioritize candidates by sequence conservation, protein properties, or regulatory features. These predictions are evidence for investigation, not direct demonstrations of an effect.

An important strategy connects SNPs to molecular phenotypes. An expression quantitative trait locus, or eQTL, is a locus associated with variation in gene expression. Related analyses examine splicing, protein abundance, DNA methylation, and chromatin accessibility. Effects are often tissue-specific or dependent on cell state. A regulatory variant that acts in an activated immune cell may show little detectable effect in an unrelated tissue or in an unstimulated cell population.

When a disease association and an eQTL occur in the same region, their overlap does not automatically mean that altered expression mediates disease risk. Distinct causal variants can be correlated through LD. Colocalization methods assess whether the association patterns are consistent with a shared underlying variant, but their conclusions depend on statistical assumptions and the available data. Establishing mediation requires additional evidence about the direction and mechanism of the relationship.

Experimental assays can test candidate mechanisms more directly. Reporter assays measure whether alternative sequences change the activity of a regulatory construct. Massively parallel reporter assays examine many sequences simultaneously, while allele-specific expression analysis compares transcripts produced from the two chromosome copies in a heterozygous individual. Each approach has limitations: a short isolated DNA fragment may not reproduce its native chromatin environment, and allele-specific measurements can be distorted by read-mapping bias.

Genome editing allows researchers to alter a candidate base in cells and observe resulting changes in expression, protein function, or cellular phenotype. Carefully controlled experiments can compare otherwise similar genetic backgrounds, helping separate a candidate variant from nearby correlated variants. Interpretation still depends on editing accuracy, cell model, off-target effects, and whether the experimental conditions capture the relevant biology. Evidence across independent assays is stronger than any single annotation or perturbation.

The lactase-persistence region illustrates this chain of reasoning. In many European populations, the variant commonly designated −13910 C>T, located in an intron of the neighboring MCM6 gene, is associated with continued expression of lactase after childhood. Regulatory evidence links the region to control of LCT expression. Other lactase-persistence-associated variants occur in some African and Middle Eastern populations. Thus, similar phenotypes can arise through different regulatory substitutions, and the nearest gene containing a SNP need not be the gene whose expression it affects.

Applications in medicine, population history, and breeding

In medicine, SNPs serve as markers of inherited susceptibility, modifiers of biological pathways, and sometimes direct contributors to clinically relevant variation. Their interpretation differs from that of highly penetrant variants responsible for some single-gene disorders. An allele can be statistically associated with disease while being neither necessary nor sufficient for disease development. Population frequency, effect size, inheritance pattern, and independent functional evidence all matter when assessing its significance.

Pharmacogenetics examines how inherited variation affects responses to medicines. Variants in CYP2C9 and the regulatory region of VKORC1 contribute to differences in warfarin dose requirements. Several SNPs in TPMT and NUDT15 affect pathways relevant to thiopurine metabolism and toxicity. These examples show how nucleotide differences can alter enzyme activity or gene regulation. Drug response nevertheless also depends on other genetic variants, age, interacting medicines, organ function, and the treatment setting.

In population research, SNPs help estimate relatedness, reconstruct demographic history, and investigate migration and admixture. Dense markers allow identification of long shared chromosome segments consistent with recent common ancestry, while allele-frequency patterns provide evidence about older population relationships. Analyses of ancient DNA connect these patterns with samples from particular archaeological periods. Such inferences are model-based and sensitive to which present-day and ancient populations have been sampled.

Ancestry estimates describe relationships to reference datasets, not fixed biological categories. A percentage assigned by an analysis depends on reference populations, marker selection, statistical methods, and the historical scale being modeled. Fine-scale geographic interpretation becomes especially uncertain where neighboring populations have exchanged genes extensively. SNP data can reveal population structure without dividing humanity into discrete, internally uniform genetic groups.

In agriculture, SNP arrays and sequencing support parentage testing, genetic mapping, and genomic selection. Breeders use large marker sets to estimate breeding values for complex traits such as milk production, disease resistance, or yield. Genomic selection differs from selecting a single favorable allele: it models many markers jointly, often exploiting their correlations with causal variants. Prediction generally works best when training and target populations are closely related and management conditions are appropriately represented.

SNPs also support conservation and species management by measuring genetic diversity, identifying population connectivity, and examining inbreeding. Marker panels developed in one population can, however, underrepresent variation elsewhere. This ascertainment bias may distort comparisons of diversity or differentiation if it is not accounted for. Whole-genome sampling can reduce some panel-design biases, but sequencing quality, reference choice, and unequal sampling remain relevant.

Interpretation limits, representation, and genetic privacy

A SNP result is a measurement embedded in several layers of inference. The underlying nucleotide call may be uncertain; the allele frequency may come from an unrepresentative sample; and a trait association may not identify a mechanism. These layers should not be conflated. A precisely measured genotype can have poorly understood biological significance, while an established association may be reported incorrectly if the wrong strand or effect allele is used.

Effect estimates require careful labeling. A risk-increasing allele in one publication may be presented as the reference or alternative allele in another, and an odds ratio below one can represent the same relationship after reversing the coded allele. Comparisons also depend on phenotype definition, covariate adjustment, and the population studied. Relative effects do not by themselves specify absolute probability: an odds ratio must be interpreted alongside baseline frequency and the model used to estimate it.

Unequal representation in genomic research limits both discovery and prediction. Many influential human association datasets have been dominated by participants of European ancestry. This imbalance affects which variants are well measured, how accurately imputation performs, and how well estimated effects transfer across populations. Broader representation can improve scientific inference, but broad labels alone do not capture within-population diversity, local demographic history, or differences in environmental exposure.

Large SNP datasets can reveal sensitive information about an individual and their biological relatives. They may support inference of kinship, ancestry, or some health-related traits, even when the original purpose of collection was different. Removing names does not make dense genetic data intrinsically anonymous, because the information is distinctive and can sometimes be linked to external records. Genetic data privacy therefore involves access controls, consent conditions, governance, and the handling of information that is partly shared among family members.

Biobanks and genetic studies also face questions about secondary use, international data transfer, commercial access, and the return of findings. Legal rules vary by jurisdiction, and protections concerning genetic discrimination do not necessarily apply uniformly across all insurance, employment, or research contexts. In studies involving communities with specific collective interests, governance may address both individual participation and community-level concerns about interpretation and use.

The scientific value of SNP information depends on preserving the distinctions among a variant, a marker, an association, and a demonstrated mechanism. Improved sequencing and larger datasets increase the number of observations, but they do not eliminate the need for reliable phenotyping, suitable comparison groups, uncertainty estimates, and experimental validation. Particularly for rare alleles, understudied populations, and context-dependent regulatory effects, the limiting factor is often not detecting the nucleotide difference but establishing what that difference does.