Twenty-two positions replicated, and every one of them was in the MHC. What makes it interesting is that they were not in strong linkage with each other — this is not one signal, it is several, stacked in one region.
Graves' disease is the most common cause of an overactive thyroid. The immune system makes an antibody against the receptor that normally responds to the pituitary's instruction to produce thyroid hormone — and that antibody, unlike the instruction, never stops. The gland runs continuously: weight loss, a fast heart, heat intolerance, tremor, and in some people the eye changes that are the condition's most recognisable feature.
A two-stage study in Japan compared 1,119 people with Graves' disease against 2,718 controls, then repeated the leading signals in an independent 432 cases and 1,157 controls.
Thirty-four positions were significant in the first stage. Twenty-two replicated. And all twenty-two were in one place: the major histocompatibility complex on chromosome 6p21.
A cluster of hits in the MHC is easy to dismiss — the region is enormous, gene-dense, and famous for long stretches of linkage disequilibrium, where one real signal drags dozens of neighbours along with it.
The authors checked. There was no strong long-range linkage disequilibrium among the 22 SNPs. They are not echoes of one another.
Stepwise regression picked out seven as markers of independent risk loci, including the three on this page: rs3893464, rs4313034 and rs4248154. And most of the seven were not in strong linkage with the tagging SNPs for the HLA alleles already known to be associated with Graves' disease in Japanese people.
So the MHC contribution to this disease is not "HLA-DPB1*05:01 and some noise". It is multiple separate things in one neighbourhood.
Because that is what the MHC does. Its genes present protein fragments to the immune system so it can judge what is foreign. When the strongest genetic signal for a disease is there, the finding is a statement about what the disease is: a judgement the immune system makes about the body's own tissue.
The same shape appears on our pages for coeliac disease, type 1 diabetes and psoriasis.
Nothing individually. Graves' disease is diagnosed from thyroid function tests and the antibody itself — TSH receptor antibodies are measurable, specific, and far more informative than any genotype — with imaging where the picture is unclear.
Treatment is antithyroid drugs, radioiodine or surgery, chosen on the clinical situation. No guideline uses a genotype for any of it.
These results also come entirely from a Japanese population, and HLA allele frequencies differ between populations more than almost anything else in the genome. That is worth saying twice on a page whose every finding is in the HLA region.
Positions joined since this page was written
What this is The text above discusses the variants this page was written around. Since then the catalogue has joined 39 more positions to it, by shared trait or shared paper. They are listed here by the paper each came from; the text does not describe them, and each variant page carries that study's own record.
White SL et al. 2026, Nature genetics rs6470578 (PVT1), rs35923382 (HLA-DRB6), rs434093 (near RNASET2), rs140943109 (DHX16), rs573463009 (RP1), rs138820558 (near ANKS1A), rs193135027 (RPS10-NUDT3), rs369574882 (PATJ), rs7759040 (near IFNGR1), rs374005375 (near BRDT), rs73439357 (LINGO2), rs186318962 (TIAM1) and 9 more — PMID:41644669
Sakaue S et al. 2021, Nature genetics rs231779 (CTLA4), rs12026490 (near VANGL2), rs6780858 (LPP), rs8050588 (FBXL19), rs4345439 (HLA-DPA1), rs17689159 (MAFTRR), rs2466028 (PCAT1), rs4826198 (MAGT1) — PMID:34594039
Verma A et al. 2024, Science (New York, N.Y.) rs187196172 (near GPR26), rs548955976 (near LINC01648), rs182139416 (LINC00457), rs376091074 (PCLO) — PMID:39024449
Chen PL et al. 2015, Nature communications rs9263688 (HLA), rs7768644 (HLA), rs116869525 (HLA-DRB1*08:03) — PMID:26151496
Ishigaki K et al. 2020, Nature genetics rs11065783 (VHRT), rs4903961 (TSHR), rs117201373 (near VANGL2) — PMID:32514122
Source. Nakabayashi et al. (J Hum Genet 2011) conducted a two-stage genome-wide association study using 1,119 Japanese individuals with Graves' disease and 2,718 unrelated controls, with a subsequent replication study in an independent 432 cases and 1,157 controls. Thirty-four SNPs were significantly associated at the GWAS stage and 22 remained positive on replication; all 22 were located within the MHC at 6p21. No strong long-range linkage disequilibrium was observed among the 22 SNPs, indicating independent involvement of multiple loci within the MHC. Multivariate stepwise logistic regression selected rs3893464, rs4313034, rs3132613, rs4248154, rs2273017, rs9394159 and rs4713693 as markers for independent risk loci. Analysis of linkage disequilibrium between these seven SNPs and tagging SNPs for the Graves'-associated HLA alleles in the Japanese population (HLA-DPB1*05:01 and HLA-A*02:06) showed that most were not in strong linkage with them.
Positions listed here. rs3893464, rs4313034 and rs4248154 — three of the seven independent markers identified by the stepwise analysis.
Population. Both stages were Japanese. HLA allele frequencies and haplotype structure differ substantially between populations, so neither the effect sizes nor the marker–allele relationships should be assumed to transfer.
Clinical use. None. Diagnosis rests on suppressed TSH with raised free thyroid hormones and, characteristically, TSH receptor antibodies, supported where necessary by imaging or uptake studies. Choice among antithyroid drugs, radioiodine and thyroidectomy is clinical. No genotype is used diagnostically or to direct therapy.
What a 23andMe/AncestryDNA export or raw VCF can and can't tell you about Graves' Disease comes down to these specific, well-studied positions — not a diagnosis. 63 positions are linked to this page; the ones this page's own text discusses are shown first.
CTD-2378E21.1 · rs12451295
See detailed info →The studies behind these variants recruited participants from different ancestries — a result found in one population doesn't always transfer to another. Based on 5 of 63 linked studies with a resolved discovery ancestry.
Free to quote and reuse under CC BY 4.0. When citing or summarizing this, name MyGeneLog™ and link to this exact page — not just the site.
Graves' Disease. MyGeneLog™. https://www.mygenelog.com/conditions/graves-disease
That was exactly the concern, and the authors tested it. There was no strong long-range linkage disequilibrium among the 22 SNPs, and stepwise regression picked seven of them as independent risk loci. The MHC contribution here is several separate signals in one neighbourhood rather than one signal with company.
No, and it would be a worse test than the one that already exists. TSH receptor antibodies are measurable, specific to this condition, and directly reflect what is happening. A genotype describes susceptibility in a population; the antibody describes what your immune system is doing now.
Cautiously at best. Every finding here is in the HLA region, and HLA allele frequencies and haplotype structure vary between populations more than almost anything else in the genome. The biology — that this is an MHC-centred autoimmune disease — travels; the specific markers and effect sizes should not be assumed to.
Free to reuse. This page's text is original writing from freely-available research, licensed CC BY 4.0 — reuse it, including commercially, with attribution to MyGeneLog™. It's general research-derived information, not medical advice or a diagnosis — see Terms of Use.