One in nine adults worldwide lives with diabetes, and the great majority of it is type 2. Hundreds of variants each nudge the risk a little — and what you eat, which your genes also have a say in, does more.
Type 2 diabetes is the condition in which the body stops responding properly to insulin and, over time, stops making enough of it. Blood sugar rises and stays high, and the damage it does — to eyes, kidneys, nerves and blood vessels — accumulates quietly over years before anything hurts.
It is common on a scale that is hard to picture. The International Diabetes Federation (IDF) counts 589 million adults aged 20 to 79 living with diabetes, about one in nine, and projects 853 million by 2050. The overwhelming majority of that is type 2.
Type 2 diabetes is polygenic to an unusual degree: hundreds of positions each shift the odds slightly, and no single one of them decides anything. The three on this site are among the better-replicated:
Put together, these three explain very little of who develops the disease. That is not a failure of the research; it is what the research found. Type 2 diabetes rose sharply in populations whose genes did not change, which by itself tells you where most of the risk lives.
Two threads run out of this page, and both are the reason it exists.
Medicines. Sulfonylureas are among the oldest diabetes drugs still in use, and CYP2C9 — the enzyme that clears them — is one of the most variable drug-metabolising genes there is. The reduced-function forms genuinely do slow that clearance. And the prescribing guidance still says not to change the dose for it. That gap between "a real biological effect" and "a reason to do something differently" is exactly what a page like this is for, and it is written up on the sulfonylureas page.
Food. Weight and diet do more to type 2 diabetes risk than every variant here combined, and what a person eats is not purely a matter of will. The bitter-taste receptor TAS2R38 decides whether a whole family of vegetables tastes sharp or of almost nothing — see bitter taste perception — and the genetics of smell and taste is where the genome most directly touches the plate. It would be wrong to draw a straight line from a taste genotype to a diabetes diagnosis, and this site will not draw one. But the honest version is still interesting: the same genome that shifts your odds a few percent also shapes what you find pleasant to eat, and the second effect is the larger one.
Nothing here diagnoses anything. Type 2 diabetes is diagnosed with a blood test — fasting glucose, HbA1c or an oral glucose tolerance test — and no genotype substitutes for one. If you have risk factors, the useful action is the blood test, not the variant. And the things that move the risk most are the unglamorous ones: weight, what you eat, how much you move, and whether it runs in your family.
Definition and diagnosis. Type 2 diabetes is a chronic disorder of glucose homeostasis characterised by insulin resistance with progressive beta-cell dysfunction. Diagnosis rests on fasting plasma glucose, HbA1c or a two-hour value on an oral glucose tolerance test, confirmed on a second occasion unless the patient is symptomatically hyperglycaemic. Genotype has no diagnostic role.
KCNQ1 rs2237892. Intronic in KCNQ1 (11p15.4), which encodes a voltage-gated potassium channel subunit. The GWAS Catalog records the C allele as the risk allele across multiple studies with odds ratios of 1.30 (p = 4 x 10-29), 1.33 (p = 1 x 10-26) and 1.45 (p = 2 x 10-42), at risk allele frequencies around 0.59 to 0.61. The locus was identified in Japanese cohorts and the effect estimates are correspondingly weighted toward East Asian ancestry; effect sizes in European-ancestry cohorts are generally smaller. Note that loss-of-function coding variants in the same gene cause long QT syndrome, an unrelated cardiac phenotype — the diabetes association is a separate, non-coding signal.
ARAP1 rs1552224. At 11q13.4, near ARAP1; reported as CENTD2 in earlier literature including the DIAGRAM meta-analyses. The A allele is the risk allele, odds ratio 1.11 to 1.14 (p = 1 x 10-22 in the largest reported analysis) at a risk allele frequency of roughly 0.81 to 0.83. The high frequency of the risk allele is worth noting when interpreting an individual result: most people carry it.
DUSP9 rs5945326. At Xq28, near DUSP9, encoding a MAP kinase phosphatase. Odds ratios of 1.14 (p = 2 x 10-12), 1.18 (p = 7 x 10-16) and 1.27 (p = 3 x 10-10) are recorded for the A allele. X-linked loci are uncommon among established type 2 diabetes signals, and the hemizygosity of male carriers complicates direct comparison of effect estimates between sexes.
Genetic architecture. Highly polygenic. Hundreds of common loci of small effect are established, and together they account for a modest fraction of heritability; polygenic scores built from them stratify populations but perform poorly as individual predictors relative to conventional risk factors. The secular rise in incidence within genetically stable populations establishes the dominant contribution of environment and lifestyle.
Pharmacogenomics. See the pgx note below. Briefly: CYP2C9 reduced-function alleles decrease clearance of several sulfonylureas, and the Dutch Pharmacogenetics Working Group nonetheless recommends no dose adjustment, a position supported by meta-analysis showing no significant excess of sulfonylurea-induced hypoglycaemia for CYP2C9*3. Metformin response has been associated with variation in SLC22A1 and ATM in some studies without reaching clinical actionability. No diabetes drug currently carries a genotype-guided dosing recommendation in routine practice.
What a 23andMe/AncestryDNA export or raw VCF can and can't tell you about Type 2 Diabetes comes down to these specific, well-studied positions — not a diagnosis.
Research-derived gene–drug associations only — not a prescription, dosing guide, or medical advice. Always follow your prescriber's guidance.
| Gene | Drug | What the research shows |
|---|---|---|
| CYP2C9 | Sulfonylureas (gliclazide, glimepiride, glibenclamide, tolbutamide) | A real metabolic effect that guidelines still say not to act on. The reduced-function CYP2C9 alleles slow the clearance of gliclazide, glimepiride and other sulfonylureas, so blood levels run higher. The Dutch Pharmacogenetics Working Group reviewed this and recommends no dose adjustment: a meta-analysis of 2,769 patients found no significant increase in sulfonylurea-induced hypoglycaemia for CYP2C9*3, and where levels do run higher the added effectiveness offsets the risk. Worth knowing precisely because the gap between "measurable in the blood" and "change what you prescribe" is where most pharmacogenomic overclaiming happens. (Dutch Pharmacogenetics Working Group (DPWG) guidance on CYP2C9 and sulfonylureas; meta-analysis of CYP2C9*3 and sulfonylurea-induced hypoglycaemia, 2,769 patients) |
No. Each of them shifts the odds by a small amount — the largest, in KCNQ1, by around 1.3 to 1.45 times per copy — and hundreds more like them exist. Most carriers never develop the disease. The rate of type 2 diabetes rose sharply in populations whose genes did not change, which tells you where most of the risk actually sits.
Because "a little" is not "nothing", and because the alternative is that you read about these variants somewhere that overstates them. The useful part is knowing the size of the effect, not just its direction.
Not currently. CYP2C9 variants really do slow the clearance of sulfonylureas, and the prescribing guidance still recommends no dose change, because the expected harm did not appear in the data. No diabetes drug carries a genotype-guided dose in routine practice.
Indirectly but genuinely: diet is one of the largest modifiable risk factors, and what a person finds pleasant to eat is partly genetic. The bitter-taste receptor TAS2R38 decides whether a family of vegetables tastes sharp or of almost nothing. That is not a line from a taste genotype to a diagnosis, and we do not draw one — but it is the same genome shaping both ends.
With a blood test — fasting glucose, HbA1c, or a two-hour oral glucose tolerance test — normally confirmed on a second occasion. No genetic result substitutes for one, and if you have risk factors the test is the useful next step.
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.