I think blue eyes are a whole lot more polygenic than this panel assumes. You might have heard of the SLC24A5, SLC45A2, MC1R, IRF4, and TYR/TYRP1/TYRP2 mutations which have complex and hard to disentangle effects across all three pigmentation domains (people will tell you MC1R and TYR(P1/2) affect just hair color, or that SLC24A5/45A2 affect just hair color, but they are wrong, changes in the melanocyte -> melanosome regulatory pathway are pleiotropic across all three domains). But recent evidence shows that the number of genes that affect these traits is even larger than just those, possibly in the domain of 10s of thousands of SNPs across 150+ genes, including exotic ones like QKI, NCOR1, LRMDA, RAB27A, BNC2, and countless others. What's more, most of these "exotic" variants underwent strong differential selection between Europeans, Africans, and Asians (see https://pmc.ncbi.nlm.nih.gov/articles/PMC10901463/). Some of them probably have their effect sizes "masked" in GWASs because they have such high linkage disequilibrium with more canonical genes – but this was not necessarily always the case, and ancient populations may have had fully depigmented phenotypes while having genotypes with none of the "canonical" mutations we focus on today, but only these weird "masked" ones.
You’re absolutely right that eye color is more polygenic and pleiotropic than just OCA2–HERC2, especially once you get into the fine-grained effects across pigmentation domains. What’s interesting, though, is that the OCA2–HERC2 haplotype (the 13-SNP h1 index I used) captures most of the signal for blue vs. brown eyes in Europeans. On its own it reaches about 90–95% predictive accuracy for those two categories, though it’s much less reliable for green/hazel or in non-European populations. Broader panels like HIrisPlex-S that add in SLC24A4, SLC45A2, TYR, IRF4, etc. can push accuracy to ~96–98%. So you’re right that the full genetic architecture is huge and still unfolding, but OCA2–HERC2 alone remains a surprisingly strong predictor. I’ll definitely have to keep an eye out for new GWAS of eye color (no pun intended) as more of these additional loci are mapped out. Which canonical genes do you think are they in LD with?
Thank you for the respectful response. I think the main HERC2/OCA2 SNPs are certainly in LD with IRF4, SLC24A5 and so forth, which might bias the results of GWAS, and also in LD with a huge array of very weakly additive (or "subtractive" i guess if the dependent variable is melanogenesis) affects, for example at LRMDA, QKI, NCOR1 etc. There was a paper a few years ago that I can't find now that looked at people with the canonical HERC2/OCA2 mutations BUT who had brown eyes, and they said the cause was most likely a set of gain-of-function mutations in TYR and TYRP1.
I think blue eyes are a whole lot more polygenic than this panel assumes. You might have heard of the SLC24A5, SLC45A2, MC1R, IRF4, and TYR/TYRP1/TYRP2 mutations which have complex and hard to disentangle effects across all three pigmentation domains (people will tell you MC1R and TYR(P1/2) affect just hair color, or that SLC24A5/45A2 affect just hair color, but they are wrong, changes in the melanocyte -> melanosome regulatory pathway are pleiotropic across all three domains). But recent evidence shows that the number of genes that affect these traits is even larger than just those, possibly in the domain of 10s of thousands of SNPs across 150+ genes, including exotic ones like QKI, NCOR1, LRMDA, RAB27A, BNC2, and countless others. What's more, most of these "exotic" variants underwent strong differential selection between Europeans, Africans, and Asians (see https://pmc.ncbi.nlm.nih.gov/articles/PMC10901463/). Some of them probably have their effect sizes "masked" in GWASs because they have such high linkage disequilibrium with more canonical genes – but this was not necessarily always the case, and ancient populations may have had fully depigmented phenotypes while having genotypes with none of the "canonical" mutations we focus on today, but only these weird "masked" ones.
You’re absolutely right that eye color is more polygenic and pleiotropic than just OCA2–HERC2, especially once you get into the fine-grained effects across pigmentation domains. What’s interesting, though, is that the OCA2–HERC2 haplotype (the 13-SNP h1 index I used) captures most of the signal for blue vs. brown eyes in Europeans. On its own it reaches about 90–95% predictive accuracy for those two categories, though it’s much less reliable for green/hazel or in non-European populations. Broader panels like HIrisPlex-S that add in SLC24A4, SLC45A2, TYR, IRF4, etc. can push accuracy to ~96–98%. So you’re right that the full genetic architecture is huge and still unfolding, but OCA2–HERC2 alone remains a surprisingly strong predictor. I’ll definitely have to keep an eye out for new GWAS of eye color (no pun intended) as more of these additional loci are mapped out. Which canonical genes do you think are they in LD with?
Thank you for the respectful response. I think the main HERC2/OCA2 SNPs are certainly in LD with IRF4, SLC24A5 and so forth, which might bias the results of GWAS, and also in LD with a huge array of very weakly additive (or "subtractive" i guess if the dependent variable is melanogenesis) affects, for example at LRMDA, QKI, NCOR1 etc. There was a paper a few years ago that I can't find now that looked at people with the canonical HERC2/OCA2 mutations BUT who had brown eyes, and they said the cause was most likely a set of gain-of-function mutations in TYR and TYRP1.