Can DNA predict eye color? The HERC2/OCA2 story

Genetics explainer · Updated June 2026

What actually colors an eye

Iris color comes down mostly to one pigment: melanin. The more melanin packed into the front layer of the iris, the browner the eye. With less melanin, more light scatters back out, and the eye looks blue — much the way the sky looks blue without actually containing blue pigment. Green and hazel sit in between, with moderate melanin plus some scattering.

Here is the part that surprises most people: there is no blue pigment in a blue eye, and no green pigment in a green eye. Human irises really only have shades of one pigment — brown-black eumelanin, with a smaller contribution from reddish-yellow pheomelanin. Every eye color you have ever seen is built from how much of that pigment sits in the iris and how the tissue behind it scatters light. A blue eye is simply an iris with very little melanin in its front layer, so short-wavelength (blue) light bounces back to the observer. Add a bit more melanin and some warmth from pheomelanin and you drift toward green or hazel. Add a lot, and incoming light is absorbed rather than scattered, and the eye reads brown.

That is why the old idea of pigments "mixing" like paint is the wrong mental model. Nothing is being blended. It is closer to a dimmer switch on a single ingredient, plus a bit of optical physics, than a palette. So the genetic question isn't "which gene paints the eye blue?" It's "which genes control how much melanin ends up in the iris?"

The HERC2/OCA2 switch

The single biggest answer is a variant called rs12913832. It lives inside the HERC2 gene, but its real job is to act as a control switch for the neighboring OCA2 gene — one of the main genes that builds melanin.

• One version of the variant keeps OCA2 active, melanin production stays high, and the eye trends brown.
• The other version turns OCA2 activity down, less melanin reaches the iris, and the eye trends blue.

In people of European ancestry, this one location explains a large fraction of the blue-versus-brown difference. That's unusually powerful — most traits are spread thinly across hundreds of genes, but here a single switch does much of the heavy lifting. It's also why blue eyes are thought to trace back to a shared ancestral change in this region.

Worth pausing on the mechanism, because it is a genuinely elegant bit of biology. rs12913832 is not inside OCA2 at all — it sits in an intron of HERC2, a stretch of DNA that doesn't code for the HERC2 protein. Instead, that region acts as a long-range enhancer: a piece of regulatory DNA that reaches over to the OCA2 gene and helps decide how strongly it is switched on. One version of the variant lets the enhancer do its job and keeps OCA2 transcription humming; the other version weakens that regulatory contact, OCA2 is expressed at lower levels, less melanin is produced in the iris, and blue eyes result. In other words, the "blue-eye" change doesn't break the melanin machinery — it just turns down the volume on a nearby gene. It is a textbook example of how a variant in non-coding DNA can have a very visible effect.

This also explains why blue eyes behave a bit like a recessive tendency without following the neat one-gene rules. To land firmly in the low-melanin, blue range, it generally helps to have the down-regulating version on both copies of the region. But because other genes are also weighing in, "generally helps" is as strong as the rule gets — and that caveat is the whole story of this trait.

Why it's a tendency, not a guarantee

Even though rs12913832 is the dominant factor, eye color is still polygenic — many genes contribute. Other pigmentation genes nudge the outcome, especially toward the intermediate shades. That's why prediction is uneven:

• Blue and brown can be called with relatively high confidence from DNA.
• Green and hazel are much harder — they emerge from combinations that don't map cleanly onto one switch.

So two people can share the same genotype at rs12913832 and still have visibly different eyes. The genetics set a strong leaning; the final color reflects the whole genetic background plus the ordinary variability of how an iris develops. This is exactly why eye color runs in families without obeying the tidy single-gene rules many of us were taught in school.

Debunking the Punnett-square myth

Most of us learned a very clean story in biology class: there's one gene for eye color, brown is dominant, blue is recessive, and a quick Punnett square tells you exactly what a child's eyes will be. It's a lovely teaching tool. It's also wrong.

Because eye color is polygenic, the single-gene inheritance model breaks down at the edges. Two blue-eyed parents can — uncommonly, but genuinely — have a brown-eyed child, and two brown-eyed parents routinely have blue-eyed children. Under the old "brown always beats blue" rule, a brown-eyed child of two blue-eyed parents was treated as basically impossible (and, unfairly, sometimes as evidence of something scandalous). Once you know that several genes contribute, the mystery evaporates: the combination the parents happen to pass on can nudge a child outside what any one switch predicts.

The honest summary is that rs12913832 loads the dice heavily, but it is still dice. Eye color inheritance follows probabilities across the whole genetic background, not a two-by-two grid.

Why eye color can change in infancy

Genotype is fixed at conception, but the visible color is not fixed at birth. Many babies of European ancestry arrive with relatively little melanin in the iris, which is why so many are born with grayish-blue eyes that later darken. Over roughly the first year of life, melanin-producing cells in the iris keep depositing pigment, and an eye that started out blue can settle into green, hazel, or brown as that pigment accumulates.

Nothing in the DNA changed — the instructions were always there. What changed is how far the iris has gotten in carrying them out. It's a nice reminder that genes describe a process unfolding over time, not a snapshot printed at birth.

"Largely but not entirely genetic"

It's fair to say eye color is largely genetic — most of the variation between people is inherited, which is why it clusters so strongly within families and ancestries. But "largely" isn't "entirely." Eye color can shift in the first year of life as melanin accumulates, and the polygenic nature means there's real spread around any prediction. Genetics gives you the odds, not a verdict.

That framing matters for every trait, not just this one: a DNA marker describes a tendency across many people with that genotype, not a fixed fact about you. rs12913832 is simply one of the clearer examples — strong enough to be genuinely predictive, but still probabilistic.

What forensic models add

Because the link is so strong, researchers built tools that predict eye color from DNA for forensic use, combining rs12913832 with a handful of other markers. This field is called forensic DNA phenotyping — inferring physical appearance from a DNA sample rather than matching it to a named person in a database. Investigators might use it when they have crime-scene DNA but no matching profile on file, to generate an investigative lead about what a person of interest may look like.

For eye color, rs12913832 does most of the work, supplemented by a small panel of additional pigmentation markers. These models do well on blue and brown and openly acknowledge that intermediate colors remain a weak spot; a responsible result is reported as a probability ("likely brown-eyed"), not a certainty. It's a real, working application of exactly the biology on this page — and a good illustration of the limits, since even a purpose-built forensic panel still hedges on green and hazel. The takeaway is consistent: more markers improve the estimate, but none of them turn a tendency into a certainty.

The bottom line

Yes — DNA can predict eye color surprisingly well, and most of that power comes from a single switch near HERC2/OCA2. But it predicts a probability, not a guarantee. Blue and brown are the easy cases; green and hazel keep geneticists humble; and the whole picture is a reminder that even a "strong" genetic marker describes a leaning, not a destiny. This page is educational, not medical advice.

Here's the fun takeaway to keep in your back pocket: the next time someone insists that two blue-eyed people can't have a brown-eyed baby, or that there's blue pigment behind blue eyes, you can gently set them straight. There's no blue paint, no green paint — just one pigment, one very clever regulatory switch near HERC2/OCA2, and a lot of light-scattering physics. Eye color is one of the most visible traits we have, and it's still humble enough to remind us that genetics deals in odds, not verdicts.

A pattern you'll see across your traits

The eye-color story is a template for how most trait markers work. If you enjoyed this, the same "one well-studied variant, big visible effect, but never the whole story" pattern shows up in other fun traits — like the redhead gene, MC1R and why cilantro tastes like soap for some people (OR6A2). None of them is deterministic, and all of them are more interesting than the tidy single-gene version you were taught.

To see which pigmentation and other markers sit in your own file, browse the Quanome gene library, try the DNA explorer (it reads your file in your browser, nothing uploaded), or read more on the Quanome blog.

Frequently asked questions

Can DNA tests predict eye color?

Reasonably well, but not perfectly. A single variant near HERC2 (rs12913832) explains a large share of the blue-versus-brown difference in people of European ancestry. Forensic models that combine several markers can predict blue or brown eyes with high accuracy, but intermediate colors like green and hazel are much harder to call.

What does rs12913832 do?

It sits in an intron of the HERC2 gene and acts as a switch that controls how active the nearby OCA2 gene is. OCA2 helps make melanin in the iris. The variant that turns OCA2 activity down is associated with less melanin and blue eyes; the other version keeps OCA2 active and is associated with brown eyes.

Why do some people's eye color not match their DNA?

Eye color is polygenic — many genes contribute, and rs12913832 is the largest but not the only factor. Other pigmentation genes, rare variants, and developmental chance all play a role, so two people with the same genotype can have noticeably different eyes.

Is eye color purely genetic?

Largely, but not entirely. Most of the variation is inherited, which is why eye color runs strongly in families. But it is a tendency shaped by many genes rather than a single guaranteed outcome, and this is educational, not medical advice.

Can two blue-eyed parents have a brown-eyed child?

Yes, though it is uncommon. The old classroom model — one gene, brown simply dominant over blue — is a simplification. Because eye color is polygenic, the combinations two parents pass on can occasionally produce a child whose eyes fall outside what a single-gene rule would predict. The reverse happens too: two brown-eyed parents can have a blue-eyed child.

Why do babies' eyes change color?

Melanin in the iris is still accumulating during the first months of life. Many babies of European ancestry are born with little iris melanin and lighter eyes, which can darken over the first year as pigment builds up. The genes were set at conception, but the visible color takes a while to settle.

Does rs12913832 predict green or hazel eyes?

Not well. rs12913832 is powerful along the blue-to-brown axis, but green, hazel, and other intermediate shades come from moderate melanin plus contributions from additional pigmentation genes. Those in-between colors are exactly where DNA prediction is weakest.

Is eye color prediction used in the real world?

Yes — forensic DNA phenotyping uses panels that include rs12913832 to estimate a person's likely eye color from a DNA sample. These tools are reasonably good at flagging blue or brown and are candid that intermediate colors remain hard to call.