Equip DNA Info Lab · Revisió editorial · 4 d’ag. del 2026 · 11 min de lectura

APOE rs429358 and rs7412: ε2, ε3 and ε4 explained

The short answer

Your APOE type is not one variant but two, read together: rs429358 and rs7412. Each chromosome carries one combination, and the pair of them is what people mean by ε2, ε3 or ε4. In the reference meta-analysis of 5,930 cases and 8,607 controls (Farrer et al., JAMA 1997), one ε4 copy carried an odds ratio of 3.2 (95% CI 2.8-3.8) and two copies 14.9 (10.8-20.6), against the common ε3/ε3, in Caucasian clinic- and autopsy-based samples. Those are odds measured between cases and controls, not your probability of developing the disease — a case-control study selects its cases and cannot estimate one. Most ε4 carriers never develop Alzheimer's.

How the two variants combine

Read one row per chromosome: a T at rs429358 with a T at rs7412 is ε2, a T with a C is ε3, a C with a C is ε4, and a C with a T is ε1 — a haplotype essentially never seen. Since you have two copies of chromosome 19, what a raw file shows you is the two genotypes, and the grid below maps every combination of them to the APOE call it is consistent with.

The middle row is the honest edge case: rs429358 CT with rs7412 CT is consistent with ε2/ε4 and with ε1/ε3. Because ε1 is so rare, this combination is in practice taken to be ε2/ε4 — but that is an inference from allele frequencies, not a reading. Distinguishing the two needs phasing or a third marker, and no tool working from these two positions alone can do it.

Allele, genotype, and risk are three different things

An allele is one letter on one chromosome. A genotype is the pair you carry. An odds ratio is a ratio measured between two groups in a study — a statistic about people who share a genotype, never a statement about you. The distinction matters as soon as you compare numbers. The 3.2 and 14.9 above are per genotype: ε3/ε4 and ε4/ε4 against ε3/ε3, in one stratum of one study. What this report computes is a different thing — the per-allele odds ratios that individual GWAS Catalog studies report for rs429358, pooled by inverse variance across the studies that measured the same trait — a variant is studied for several, and pooling across them would attach one number to the name of one of them — each study weighted by 1/SE², so a precise one counts for more than an imprecise one — and used as an internal heuristic weight rather than a validated estimate. The median is only a fallback, for the variants where no study reported an interval. It restricts the pool to European-ancestry studies, which is both the improvement and the limitation: it stops mixing strata that disagree, and we do not know your ancestry. It still does not preserve age, sex or study design, and this category is now restricted to studies of Alzheimer disease itself — dementia and cognitive decline used to be folded in with it, with individual values running from below 1 (protective) to above 4. Figures from different designs and different populations are not interchangeable, and none of them is your personal probability or a clinical estimate of your risk.

Two things that silently change the letters

Strand. rs429358 is reported as T/C on one strand and as A/G on the other. A file showing AG where you expected TC may be the same result written on the complementary strand — not a different genotype. Check the dbSNP record before concluding anything.

Build. The same position is 19:45411941 on GRCh37 (hg19), which is the build the 23andMe, AncestryDNA and MyHeritage exports this site parses are written in, and 19:44908684 on GRCh38. Comparing a coordinate from one build against a file in the other is the most common way to conclude a variant is "missing" when it is present.

If you've ever read about consumer DNA testing, you've probably seen the variant rs429358 mentioned. It sits in the APOE gene on chromosome 19 and, together with a second variant rs7412, defines the famous APOE ε2/ε3/ε4 alleles. One of those — ε4 — is the strongest common genetic risk factor for late-onset Alzheimer's disease known to science.

What rs429358 is, in one paragraph

APOE codes for apolipoprotein E, a protein that carries cholesterol in the brain. At position rs429358, you carry either a T or a C. A T at that position contributes to the ε2 or ε3 allele; a C contributes to ε4. Since you have two copies of chromosome 19, your genotype is two letters — TT, CT or CC — and the combination with rs7412 determines whether you carry zero, one or two copies of the ε4 allele.

How much those numbers vary

The 3.2 and 14.9 above are one stratum of one meta-analysis, and the same paper shows how far the effect moves. In Japanese subjects it was larger, not smaller: 5.6 (3.9-8.0) for ε3/ε4 and 33.1 (13.6-80.5) for ε4/ε4. In African American and Hispanic samples the ε4 association was notably weaker, with significant heterogeneity between the African American studies. ε2 goes the other way — ε2/ε3 carried an odds ratio of 0.6 (0.5-0.8), so the grid above is worth reading in both directions. And the ε4 effect is present between ages 40 and 90 but diminishes after 70. Any single multiplier quoted for ε4 is one of these numbers with its stratum removed.

What it doesn't tell you

A high-risk APOE genotype is not a diagnosis or a prediction. The majority of ε4 carriers never develop Alzheimer's, and many non-carriers do. Risk is also modulated by other genes (TREM2, CLU, PICALM, dozens more), age, cardiovascular health, education, sleep and lifestyle. APOE explains a meaningful slice of the variation between people, not the outcome for any one person.

And it does not give you a number that transfers. As the figures above show, the size of the association varies with ancestry, age, sex and study design, and it does not vary in one direction — larger in the Japanese samples, weaker in the African American and Hispanic ones. Most of the foundational APOE work was done in European-ancestry populations, so that is the stratum the familiar multipliers come from; applying it to someone else is an assumption, not a result.

How to check yours

If you already have a raw DNA file from 23andMe or AncestryDNA, you can upload it and rs429358 will appear in the Alzheimer's section of your report, with its odds ratio in plain language and links to the peer-reviewed studies behind it. Three things to know first. We do not derive an ε2/ε3/ε4 call: the report reports each variant on its own, so it will not tell you which APOE alleles you carry. And rs429358 is not on every chip version — if your file does not contain it, the Alzheimer's section is built without its best-studied variant, and the report will not warn you. Its coverage figure drops by one position out of hundreds, which is a fact about your file and not a judgement: coverage counts how many studied positions are present, it does not weigh how important a missing one is, and it is not a reliability warning. And the report no longer shows an overall score, band or percentile for any category: our own audit found that most of a category’s combined effect came from studies measuring something other than the condition it names, so we withdrew the number rather than publish one we cannot defend. What it shows is each variant with the trait its study measured and a link to it — so read the APOE result as the association described above, not as a personal multiplier.

If you'd rather not know your APOE status, that's a legitimate choice — many clinicians actively discourage testing without genetic counseling because the psychological impact of an ε4/ε4 result is real. Our report shows the result only inside a collapsible card that you can choose not to open.

Further reading

• Polygenic risk scores (PRS) for non-scientists → — APOE ε4 is one strong effect carried by two variants together, but most disease risk is polygenic.
• Our methodology → — how we read GWAS data, and why we publish no overall score.

I com es veu rs429358 dins del fitxer? La fila de rs429358 en un exemple anotat →

This article is for educational purposes only and does not constitute medical advice. If you have concerns about Alzheimer's risk, please consult a healthcare professional or genetic counselor.