Reading your APOE genotype

APOE encodes apolipoprotein E, a 299-amino-acid protein produced primarily by astrocytes and microglia in the central nervous system, and by hepatocytes in the periphery. APOE participates in lipid transport, and in the brain it is the principal carrier of cholesterol between glia and neurons. The three common alleles (ε2, ε3, ε4) differ at two amino-acid positions (residues 112 and 158), which alter the protein's receptor binding and its interaction with amyloid-beta peptides.

Most people carry one or two copies of ε3, the reference allele. About one in four people of European ancestry carries at least one copy of ε4, and that allele is the strongest known common-variant risk factor for late-onset Alzheimer's disease. The mechanism is not fully resolved, but converging evidence implicates reduced microglial clearance of amyloid-beta, altered tau processing, weaker blood-brain-barrier integrity, and lipid-handling abnormalities. Each mechanism modestly increases risk; together they explain roughly 25% of the population variance in late-onset AD risk.

The genotype table

• ε2/ε2: rare and slightly protective against Alzheimer's, but associated with a small risk of type III hyperlipoproteinaemia in middle age.
• ε2/ε3: slightly reduced Alzheimer's risk relative to ε3/ε3.
• ε3/ε3: the population reference. Roughly two thirds of Europeans carry this combination.
• ε3/ε4: about a 2 to 3-fold increased lifetime risk of late-onset Alzheimer's.
• ε2/ε4: roughly background risk; the protective ε2 partially offsets ε4.
• ε4/ε4: about a 10 to 15-fold increased lifetime risk; estimated lifetime risk of around 50 to 60% by age 85 in some cohorts.

Why we resolve APOE in four tiers

APOE genotyping looks easy in principle because only two single-nucleotide polymorphisms (rs429358 and rs7412) define the ε2/ε3/ε4 alleles. In practice, every consumer chip is different. Older 23andMe versions encode these positions under proprietary identifiers (i3003137 and i3002555) that some downstream tools fail to map. WGS and microarray data sometimes report the variants by chromosomal position alone, without an rsID at all. Many imputed datasets carry only proxy variants in linkage disequilibrium with the true causal SNPs.

Haeckel works through a four-tier resolution cascade so the call rate stays high without sacrificing accuracy.

1. Tier 1: read rs429358 and rs7412 directly when available. Highest confidence.
2. Tier 2: fall back to the 23andMe internal identifiers i3003137 and i3002555 if the rsIDs are absent. Same biology, different label.
3. Tier 3: match by chromosomal position (chr19 positions 44908684 and 44908822 in GRCh38) when no identifier matches.
4. Tier 4: estimate from a weighted vote across at least two of three proxy SNPs (rs440446, rs405509 and rs439401). Lower confidence, but better than refusing to call.

Each result reports both the inferred genotype and the tier it came from, so you can see exactly how much trust to place in the call. The proxy SNPs used in Tier 4 are rs440446, rs405509, and rs439401. Their reported r² values with the canonical pair are around 0.65, 0.40, and 0.35 respectively, which means a Tier 4 call is informative but should not be treated as a clinical-grade APOE call. Tier 4 calls explicitly recommend confirmatory testing if the result would change a clinical decision.

Sex and ethnicity modify the association

The APOE-Alzheimer's association is not uniform across populations. ε4 carriers among Yoruba and Hispanic cohorts show smaller risk increases than European ε4 carriers (Farrer 1997, JAMA), in part because the population background risk differs. Within any given population, the association is stronger in women than in men: a female ε3/ε4 carrier carries roughly 1.5 to 2 times the risk increase that a male ε3/ε4 carrier carries, an effect that becomes prominent after age 65 and that is hypothesised to relate to oestrogen signalling and microglial activation. The risk numbers in the genotype table above are population-averaged; the per-individual figures Mirror surfaces are calibrated against your specific ancestry and sex when both are available.

Risk modifiers worth knowing about

• TOMM40: a gene immediately adjacent to APOE that is in strong linkage disequilibrium with the APOE alleles. Some TOMM40 variants modify the age of dementia onset by 5 to 7 years even after controlling for APOE.
• Complement system genes (CR1, C3, CLU): each contributes a small additional effect, typically on the order of 1.1 to 1.2-fold risk per allele copy. Combined into a polygenic score, they explain a few additional percent of the AD risk variance.
• Education and cognitive reserve: documented in many cohorts to modify the age of clinical onset. The effect is real but does not change the underlying neuropathology.
• Cardiovascular health: the FINGER and Lancet Commission analyses both find that managing midlife hypertension, diabetes, and lipids reduces dementia risk in ε4 carriers more than in non-carriers.

Lifestyle interventions with the strongest evidence

• Physical exercise (FINGER trial, 2015): structured aerobic plus resistance training in midlife shows the strongest single intervention effect on cognitive trajectory in ε4 carriers.
• Mediterranean diet (MIND diet): consistent observational evidence and several controlled trials suggest protection. The signal is stronger when adherence is sustained over decades.
• Sleep quality: deep sleep clears amyloid-beta through the glymphatic system. Chronic sleep deprivation in ε4 carriers shows a measurable amyloid burden increase on PET imaging.
• Hearing aids in mid-to-late life: the 2024 Lancet Commission update rated untreated hearing loss as the largest modifiable midlife dementia risk factor. Treating hearing loss reduces cognitive decline rate in ε4 carriers in particular.
• Smoking cessation, alcohol moderation, and avoidance of head injury: standard population-level recommendations, with proportionally larger effect sizes in ε4 carriers.