Links from this page to related variants, conditions, drugs and topics are not included for automated readers. Those connections are the Gene Knowledge Graph and are licensed as data: https://www.mygenelog.com/license
One position explains most of whether eyes are blue or brown — which is why the brown-dominant, blue-recessive square from school is wrong. The same genes that shade normal color cause albinism when they break outright.
Eye color, skin color and hair color are the same story told three times. All of them come down to melanin: how much a cell makes, which kind, and where it ends up. The pigment is built inside melanosomes, small compartments in a melanocyte, and it starts from a single amino acid — tyrosine — which the enzyme tyrosinase converts step by step into the polymer we see as color.
An iris is not painted. Its color is what is left after melanin in the front layer absorbs some light and the tissue behind scatters the rest. Plenty of melanin absorbs nearly everything and the eye reads brown. Very little melanin absorbs almost nothing, scattering sends the short wavelengths back, and the eye reads blue — the same physics that makes the sky blue, not a blue pigment. Nobody has blue pigment in their eyes.
Brown dominant, blue recessive, two blue-eyed parents can never have a brown-eyed child. It is the classic exercise, and it does not survive contact with the data. Eye color is quantitative and several genes contribute, so blue-eyed parents can have a brown-eyed child. It is uncommon, but it happens and it is not evidence of anything except that the school model was a simplification.
What makes the myth so durable is that one position gets you unusually far. A 2008 study found that rs12913832 alone predicted eye color with an R² of 0.68 — better than the best combination of nearby variants known before it — with an association strength of LOD = 444. For a visible human trait that is extraordinary; almost nothing else on this site comes close. But 0.68 is not 1.0, and the missing third is exactly where green, hazel and every argument about a family photograph lives.
Here is the part worth slowing down for. rs12913832 is not in the pigment gene. It sits inside intron 86 of a neighbouring gene, HERC2, about 21,000 bases upstream of the promoter of OCA2 — the gene that matters. That stretch of DNA is a switch: a laboratory assay showed the element turns the OCA2 promoter down, and the two versions of the position bind different sets of proteins. The blue-associated version breaks a conserved binding site, so less OCA2 gets made in the melanocytes of the iris, so less melanin is made there.
So the variant does not change a protein at all. It changes how loudly a gene is switched on, in one tissue. That is what most of the genome does, and this is one of the few cases where the consequence is something you can see in a mirror.
The 2008 work also found the blue haplotype in blue-eyed people in Denmark, Turkey and Jordan alike — one version, spread from a single ancestor, rather than the same change happening repeatedly.
These are shades, not switches. Each moves color a little, and they are why two people with the same rs12913832 genotype are not the same color.
This is the part that matters most and gets confused most. TYR and OCA2 appear on this page as common positions that shade normal color. The same genes, when both copies carry a rare variant that breaks them, cause oculocutaneous albinism — reduced or absent melanin in skin, hair and eyes, together with real ocular effects on vision. That is a recessive condition with at least 23 genes known to cause it, and nothing on this page tests for it or hints at it.
Light eyes are not mild albinism. They are the ordinary end of ordinary variation.
It is not medical. No guideline uses eye color genetics for anything, and there is no decision here to make. It is used in forensic DNA phenotyping, where investigators predict appearance from a sample with no suspect, and it works best at exactly the place it works best generally: telling blue from brown, and much less well at everything in between.
The one practical thread is sun, and it does not need a genotype. Less melanin means less protection from ultraviolet light, which is why fair skin burns faster. That is visible without testing, and the response — cover up, use sunscreen, keep an eye on moles — is the same advice for everyone regardless of what any of these positions say.
2026-01-01 · Genome-wide association study reveals genetic architecture and evolution of human retinal pigmentation. Science Advances. 2026. DOI:10.1126/sciadv.adw7768
Retinal pigmentation -- how much melanin lines the back of the eye, protecting it from light-induced damage -- had not been studied at genome-wide scale, largely for lack of a way to quantify it from routine eye photos. This study built a deep learning tool (DeepGRP) to score retinal pigmentation from fundus images, then ran a GWAS finding 42 signals, 26 of them newly identified, with SNP-based heritability of 21.4%. Single-cell chromatin and RNA sequencing of human fetal retinal tissue pointed to retinal pigment epithelium and photoreceptor cells as the key cell types involved, and identified ARHGAP18 as a previously unrecognized regulator of melanin production (melanogenesis). The study found evidence of polygenic adaptation specifically in Europeans, consistent with selection pressure from snow-reflected light at high latitudes -- a genuine evolutionary genetics finding, not just a disease-risk one. Most practically, a polygenic score for retinal pigmentation correlated with a 4.8-fold higher risk of myopia and, in the opposite direction, a 1.5-fold lower risk of skin cancer -- a real genetic tradeoff between two very different outcomes tied to the same pigmentation biology. This site's eye color and pigmentation page carries just 4 variants; ARHGAP18 and the other 41 signals are not currently among them. Given the skin cancer connection, this site's sun sensitivity and skin cancer risk page is also a genuine match.
Chemistry. Melanogenesis begins with tyrosine. Tyrosinase (TYR), a copper-dependent enzyme, hydroxylates tyrosine to L-DOPA and oxidises L-DOPA to dopaquinone; that step is rate-limiting for the whole pathway. Dopaquinone then branches: without a thiol it cyclises and polymerises toward eumelanin (brown-black), while in the presence of cysteine it forms cysteinyl-DOPA and proceeds toward pheomelanin (red-yellow). The branch point, not the total, is why red hair and dark hair are chemically different products rather than different amounts of one product. All of it happens inside the melanosome, whose internal pH and ion content govern tyrosinase activity — which is where transporters such as OCA2 and SLC45A2 exert their effect, by setting the environment the enzyme works in rather than by catalysing anything themselves.
rs12913832 (HERC2/OCA2). A regulatory variant in intron 86 of HERC2, roughly 21.2 kb upstream of the OCA2 promoter. Reporter assays show the element reduces OCA2 promoter activity, and the two alleles bind different nuclear factors; the blue-associated allele disrupts a conserved consensus site for the helicase-like transcription factor HLTF. In a European sample the blue-associated allele frequency was 78%, and ordinal logistic regression on eye color gave R² = 0.68 with LOD = 444, outperforming prior OCA2 haplotypes. OCA2 R419Q (rs1800407) modifies penetrance. Earlier linkage work localised the signal to a 166 kb region within HERC2 and found a single haplotype shared by blue-eyed individuals from Denmark, Turkey and Jordan, consistent with one founder rather than recurrent mutation.
TYR and MATP/SLC45A2 rows. rs1393350 and rs1847134 are recorded against TYR from a large web-recruited cohort that replicated the established pigmentation loci — OCA2, HERC2, SLC45A2, SLC24A4, IRF4, TYR, TYRP1, ASIP, MC1R — across hair color, eye color and freckling. rs35391 is recorded against MATP for tanning response, from a multistage GWAS of over 9,000 people of European ancestry; that study's own novel candidate, rs966321, reached genome-wide significance in the discovery stage and then failed to replicate in two further samples, which is the ordinary fate of a discovery-stage hit and the reason replication is the bar this site holds to.
Population structure. Pigmentation loci vary sharply between ancestries, and effect estimates derived in European-ancestry cohorts do not transfer. rs12913832 discriminates blue from brown well in populations where blue eyes are common and says much less elsewhere, where brown eyes are near-universal for reasons that have nothing to do with this position.
Boundary with disease. Biallelic loss-of-function variants in TYR, OCA2 and SLC45A2 — among at least 23 genes now associated — cause oculocutaneous albinism, a recessive disorder combining hypopigmentation of skin, hair and eyes with ocular abnormalities including reduced visual acuity, nystagmus and foveal hypoplasia; syndromic forms (Hermansky-Pudlak, Chediak-Higashi) add haematological and immunological features. Current molecular diagnostic yield is about 70%. None of the common variants on this page is diagnostic of, or a carrier test for, any of that.
The studies behind these variants recruited participants from different ancestries — a result found in one population doesn't always transfer to another. Based on 4 of 61 linked studies with a resolved discovery ancestry.
Licensed CC BY-NC 4.0: free to read, share and adapt for non-commercial use, naming MyGeneLog™ and linking the exact page. Commercial use, including training or serving a commercial AI system, needs a licence — https://www.mygenelog.com/license
Eye Color and Pigmentation. MyGeneLog™. https://www.mygenelog.com/conditions/eye-color-and-pigmentation
Yes. It is uncommon, and it is not a surprise to geneticists. Eye color is quantitative and several genes contribute, so the dominant/recessive square taught in school is a simplification rather than a rule. One position gets you most of the way — about two-thirds of the variation — and the remaining third is exactly where cases like this live.
It is a common inherited version of one position, found on a single haplotype in blue-eyed people from Denmark to Jordan, which points to one ancestor rather than the same change happening again and again. Calling it a mutation is technically fine and misleading in ordinary use: it is not a defect and it does nothing but turn one gene down in one tissue.
No, and the distinction matters. Albinism is recessive and caused by rare variants that break these genes outright, with real effects on vision. The variants here are common versions that shade normal color and affect nothing else. Nothing on this page is a test for albinism or for carrying it.
Better than chance and not reliably. Blue versus brown is predicted well; green and hazel are predicted poorly, and prediction built in one ancestry does not carry to another. It is also a trait you can simply look at in a year or so, which is the more reliable test.
Not by itself. Less melanin means less protection from ultraviolet light, which is why fair skin burns faster — but that is visible without a test, and the response is the same for everyone: cover up, use sunscreen, and have changing moles looked at.
Free to reuse, non-commercially. This page's text is original writing from freely-available research, licensed CC BY-NC 4.0 — share and adapt it with attribution to MyGeneLog™; commercial use is licensed. It's general research-derived information, not medical advice or a diagnosis — see Terms of Use.