Single-Nucleotide Polymorphism

A single-nucleotide polymorphism (SNP) is a germline point mutation consisting of a single base-pair substitution (A, C, G, or T) at a specific genomic coordinate that segregates within a biological population. Classically in human population genetics, a site is designated a SNP when the minor allele frequency (MAF)—the frequency of the second most common allele in a reference population—equals or exceeds 1% (MAF ≥ 0.01), whereas rarer single-base substitutions (MAF < 0.01) or somatic mutations in tumors are encompassed by the broader term single-nucleotide variant (SNV). (Note that single-base insertions or deletions are classified separately as indels, not SNPs.)

Across the 3.05 × 10⁹ base pairs of the haploid human reference genome, SNPs represent the most abundant class of heritable genetic variation. Any two unrelated human haploid genomes differ at an average of ~1 in every 1,000 base pairs (nucleotide diversity π ≈ 0.001, corresponding to ~3.5 × 10⁶ to 4.5 × 10⁶ SNPs per individual diploid genome). Cataloged across global sequencing consortia including the 1000 Genomes Project, the NCBI Short Genetic Variations database (dbSNP), and the Genome Aggregation Database (gnomAD), more than 6.0 × 10⁸ validated human SNVs and common SNPs have been mapped, providing high-density markers for genome-wide association studies (GWAS), Mendelian randomization, population demography (genetic drift and natural selection), and clinical pharmacogenomics.

Chemically, single-nucleotide substitutions are partitioned into two structural classes based on the heterocyclic ring geometry of the substituted nitrogenous bases in DNA:

• Transitions (Ti): Substitutions that exchange a two-ring purine for another purine (A ↔ G) or a single-ring pyrimidine for another pyrimidine (C ↔ T). There are four possible transition pathways.
• Transversions (Tv): Substitutions that exchange a purine for a pyrimidine or vice versa (A ↔ C, A ↔ T, G ↔ C, G ↔ T). There are eight possible transversion pathways.

If point mutations occurred purely at random with equal probability across all twelve substitution paths, the expected ratio of transitions to transversions (Ti/Tv ratio) would be 4 / 8 = 0.50. Empirically, however, human whole-genome sequencing consistently yields a genome-wide Ti/Tv ratio of ~2.0 to 2.1, and an even higher Ti/Tv ratio of ~2.8 to 3.2 within protein-coding exons (a metric routinely used as a quality-control filter in variant-calling pipelines to detect false-positive sequencing artifacts).

Two biochemical and evolutionary mechanisms drive this four- to six-fold enrichment of transitions over the random expectation:

1. Spontaneous Hydrolytic Deamination of 5-Methylcytosine at CpG Dinucleotides: In vertebrate genomes, 70% to 80% of cytosine residues within 5′-CpG-3′ dinucleotides are methylated at carbon-5 (5mC) by DNA methyltransferases. Whereas unmethylated cytosine deaminates spontaneously into uracil (which is recognized as foreign to DNA and excised by uracil-DNA glycosylase), hydrolytic deamination of 5-methylcytosine removes the C4 amino group to generate thymine (5mC → T), a canonical DNA base. If the resulting T:G mismatch is not repaired by methyl-CpG-binding domain protein 4 (MBD4) or thymine-DNA glycosylase (TDG) prior to DNA replication, one daughter strand fixes a permanent C → T transition (or a G → A transition on the antisense strand). Consequently, CpG dinucleotides mutate at a rate ~10- to 50-fold higher than non-CpG sites (~1.2 × 10⁻⁷ vs. ~1.2 × 10⁻⁸ per base pair per generation).
2. Polymerase Geometry and Genetic Code Degeneracy: During replication, purine-pyrimidine G:T wobble mispairs distort the double-helix minor groove less severely than purine-purine (A:G) or pyrimidine-pyrimidine (C:T) steric clashes, escaping exonucleolytic proofreading more frequently. Furthermore, within protein-coding sequences, transitions at the third position of codons ( wobble position) are nearly always synonymous (silent), whereas transversions more frequently alter the encoded amino acid and are eliminated from the population by purifying selection, elevating exonic Ti/Tv to ~3.0.

Because only ~1.2% to 1.5% of the human genome encodes protein exons, roughly 98.5% of SNPs fall within intergenic regions, introns, or untranslated regions (5′-UTR and 3′-UTR). SNPs are classified functionally by their impact on transcription, RNA processing, and translation:

Alleles at physically adjacent SNP loci along a chromosome do not segregate independently at the population level; instead, they co-segregate on ancestral chromosome segments termed haplotypes that are broken up over generations only at discrete meiotic recombination hotspots (often directed by the histone methyltransferase PRDM9). The non-random statistical association between alleles at two loci (A/a with allele frequencies p_A, p_a and B/b with frequencies p_B, p_b) is quantified by the linkage disequilibrium (LD) coefficient D:

D = p_AB − (p_A × p_B)

where p_AB is the observed population frequency of the AB gametic haplotype. Because the numerical range of D depends on allele frequencies, two normalized LD metrics are standard in genomics:

1. Lewontin's D′ (|D| / D_max): Ranges from 0 to 1. A value of *D*′ = 1.0 indicates complete LD, meaning that no historical meiotic recombination has occurred between the two SNPs since the younger mutation arose (at least one of the four possible haplotypes is absent).
2. Pearson's Squared Correlation Coefficient (r²): Defined as r² = D² / (p_A × p_a × p_B × p_b). Ranging from 0 to 1, r² equals 1.0 (perfect LD) only when two SNPs have identical allele frequencies and co-segregate without recombination.

In Genome-Wide Association Studies (GWAS), the statistical power required to detect a causal variant via an indirect tag-SNP scales directly with r² (specifically, a sample size of N / r² individuals is required at the tag-SNP to achieve the same power as genotyping N individuals directly at the causal SNP). Because the human genome is structured into 10 to 50 kb haplotype blocks of high LD (r² > 0.8), genotyping arrays assaying ~500,000 to 1,000,000 tag-SNPs can impute >90% of common variation across the genome. To control the family-wise error rate across ~10⁶ independent haplotype tests in the human genome, GWAS apply a Bonferroni-corrected genome-wide significance threshold of P < 5 × 10⁻⁸ (0.05 / 10⁶).

In clinical pharmacogenomics, common coding and regulatory SNPs in phase I cytochrome P450 enzymes and phase II transferases govern inter-individual variability in drug clearance and toxicity. For example, loss-of-function SNPs in CYP2C19 (*2 splice defect rs4244285 and *3 nonsense variant rs4986893) prevent bioactivation of the antiplatelet prodrug clopidogrel, whereas promoter and coding SNPs in VKORC1 (rs9923231, −1639G>A) and CYP2C9 (*2 and *3) account for ~40% to 50% of dose variance for the anticoagulant warfarin.

In evolutionary genomics, the ratio of nonsynonymous substitutions per nonsynonymous site to synonymous substitutions per synonymous site (ω = d_N/d_S) and the site frequency spectrum (Tajima's D, extended haplotype homozygosity EHH, and iHS) distinguish neutral polymorphism from adaptive sweeps: selective sweeps driven by positive selection—such as the EPAS1 hypoxia-adaptation haplotype in Tibetan populations or the DARC (ACKR1 rs2814778, −67T>C) Duffy-null promoter SNP conferring resistance to Plasmodium vivax malaria across sub-Saharan Africa—generate long, high-frequency haplotype blocks with locally depleted SNP diversity.

1. The 1000 Genomes Project Consortium. (2015). A global reference for human genetic variation. Nature, 526(7571), 68–74. https://doi.org/10.1038/nature15393
2. Slatkin, M. (2008). Linkage disequilibrium — understanding the evolutionary past and mapping the medical future. Nature Reviews Genetics, 9(6), 477–485. https://doi.org/10.1038/nrg2361
3. Karczewski, K. J., Francioli, L. C., Tiao, G., Cummings, B. B., Alföldi, J., Wang, Q., Collins, R. L., Laricchia, K. M., Ganna, A., Birnbaum, D. P., Gauthier, L. D., Brand, H., Solomonson, M., Watts, N. A., Rhodes, D., Singer-Berk, M., England, E. M., Seaby, E. G., Kosmicki, J. A., … MacArthur, D. G. (2020). The mutational constraint spectrum quantified from variation in 141,456 humans. Nature, 581(7809), 434–443. https://doi.org/10.1038/s41586-020-2308-7
4. Relling, M. V., & Evans, W. E. (2015). Pharmacogenomics in the clinic. Nature, 526(7573), 343–350. https://doi.org/10.1038/nature15817