Alzheimer's disease

Late-onset Alzheimer's risk is partly genetic, with APOE the largest single piece. This score combines APOE and other common variants into one estimate.

What this Alzheimer's score reads

This is PGS000779, a seven-variant score built by Zhou and colleagues in 2020 for Alzheimer's disease risk in a Chinese population, using LASSO regression over variants from a GWAS of 147,484 people that was about 91% European ancestry.

Seven variants is not a polygenic score in the usual sense. Four of them sit in the same stretch of chromosome 19 around APOE, in TOMM40, NECTIN2 and APOE itself, and those four carry the largest weights. In practice this is close to an APOE result with a few additions, one of which is a protective variant near SORL1.

How well it predicts Alzheimer's disease

In a European-ancestry validation sample of 2,696 people (464 with Alzheimer's disease), the score reached an area under the curve of 0.717 with age and sex in the model. In the study's own East Asian sample the area under the curve was 0.612, though that sample held only 112 people and 33 cases, which makes the estimate imprecise.

Age remains the dominant risk factor for late-onset Alzheimer's disease by a wide margin, and a large part of the discrimination in any such model comes from age rather than genetics.

Read the APOE page instead, or as well

Because this score is largely a reflection of the APOE region, the more direct and better-established answer is the APOE result itself, which this site reads on its own page. There you see your e2, e3 and e4 combination and what each pairing means, which is how the evidence on APOE and Alzheimer's is actually reported.

A percentile that blends APOE with correlated neighbouring markers is harder to interpret than the genotype it mostly reflects, and it can give the impression of independent information that is not there.

What this score does not capture

Early-onset autosomal dominant Alzheimer's disease, caused by variants in APP, PSEN1 or PSEN2, is a different condition genetically and is not read here or by a consumer chip. Dementia appearing in a family before 60, across generations, is a reason for clinical genetics input rather than a polygenic score.

Larger studies since this score was built have identified dozens more loci, so a seven-variant score is out of date as a summary of common-variant risk even setting aside its APOE concentration.

A note on what this information does

There is no treatment that clearly changes the course of late-onset Alzheimer's disease on the basis of a genetic result, and the American College of Medical Genetics is explicit that a polygenic score is a statistical prediction, not a diagnosis, and that a low score does not rule out risk. Cardiovascular health, physical activity, hearing correction, sleep and staying socially and mentally active are the levers with evidence behind them, and they apply regardless of genotype.

Some people find this information useful for planning, and others find it distressing without being actionable. Genetic counselling exists for exactly this kind of result, and is worth considering before reading further if Alzheimer's disease runs in your family.

Variants in this score you can read on their own

This score includes a variant that this site also explains individually, where you can see what each genotype means on its own: APOE type (Alzheimer's & cholesterol) (rs429358).

How much of this score your file covers

All 7 variants in the published score are present in a standard consumer DNA file, so this reconstructs the published score in full.

Common questions

Is this the same as an APOE test? Nearly, and less clearly. Four of the seven variants lie in the APOE region and carry the largest weights, so the score mostly reflects APOE while being harder to interpret than the e2, e3 and e4 genotype itself, which this site reports on its own page.

Does a high score mean I will develop Alzheimer's disease? No. Even APOE e4 homozygosity, the strongest common genetic risk factor, leaves many people unaffected, and this score reached an area under the curve of 0.717 including age. It is a statistical prediction about groups, not a forecast for a person.

Related

Autoimmune protection (IL23R R381Q) · Coronary artery disease · Depression · Diastolic blood pressure · Fasting glucose / type 2 diabetes (MTNR1B) · HbA1c (blood sugar) · Venous thromboembolism (clots)