Alzheimer genetics comes in three tiers that behave nothing alike: rare changes that nearly decide the outcome, one common variant that matters a great deal, and a long tail that each move risk a little. This catalogue holds the second and part of the third, and none of the first — and several of the famous genes are here under completely different headings.
Solid lines are connections this site curates. Dashed lines mean the two ends share a research paper — worth knowing, and not a claim that one explains the other.
Almost everything confusing about the genetics of Alzheimer's disease comes from treating it as one thing. It is three, and they behave nothing alike.
More than 200 highly penetrant pathogenic variants in three genes — APP, PSEN1 and PSEN2 — cause a subset of early-onset familial Alzheimer's disease. Carrying one is close to determinative, and the disease usually begins well before the age most people associate with it.
This site holds none of them, and could not. They are rare, they are found by sequencing affected families, and everything in this catalogue comes from studies that scan for common differences between large groups of people. A method built to find what many people share cannot see what a few families carry. If early-onset dementia runs in your family, that is a conversation with a genetic counsellor, not a page to read.
APOE is the strongest common genetic influence on late-onset Alzheimer's disease, large enough that it is worth its own page. That page is here, including what the e2, e3 and e4 forms do and why a genotype is a probability rather than a verdict.
Beyond APOE, more than two dozen further genes have been associated with late-onset disease, each shifting risk slightly. A 2011 consortium study named MS4A4/MS4A6E, CD2AP, CD33 and EPHA1, and confirmed earlier findings at CR1, CLU and BIN1.
The heritability of Alzheimer's disease is estimated to be as high as 80%. Add up every tier above and a large share of that is still unaccounted for.
A 2019 study took a different approach to the same tier-three problem: instead of only recruiting people with a clinical Alzheimer's diagnosis, it added "AD-by-proxy" cases — people who do not themselves have a diagnosis, but whose parent does, identified from UK Biobank family-history questions. That let the study reach 71,880 cases and 383,378 controls, far larger than a clinically-diagnosed-only sample could realistically assemble. The proxy approach only works because it was checked first: AD-by-proxy status correlates strongly with clinically diagnosed AD (genetic correlation rg = 0.81), which is the paper's own stated justification for combining the two groups into one analysis.
That larger sample found 29 risk loci implicating 215 candidate genes, concentrated in immune-related tissue and cell types — spleen, liver, and microglia, the brain's resident immune cells. Gene-set analysis pointed at two mechanisms: lipid-handling pathways, and degradation of the amyloid precursor protein itself. A separate Mendelian-randomization analysis in the same paper found that higher cognitive ability appears protective against AD risk.
Under an Alzheimer trait, this site holds rs2075650 in APOE, rs7561528 in BIN1, rs115550680 in ABCA7, rs9349407 in CD2AP, rs10498633 in SLC24A4, rs6857 in the PVRL2 region, and — from the 2019 proxy-case study — rs187370608 in TREM2 and rs867611 in PICALM, two genes long associated with Alzheimer's disease that had no variant on this site at all until this study landed.
Now the part that says what a catalogue is. Several genes famous for Alzheimer's disease are in this site, filed under something else entirely — and, for two of them, under something else as well as directly under this page:
| Gene | Known for | What it is filed under here |
|---|---|---|
| CLU | a replicated late-onset Alzheimer locus | Schizophrenia |
| SORL1 | amyloid precursor sorting | Atrial fibrillation, and Alcohol consumption |
| MS4A4E | the MS4A Alzheimer cluster | Lipoprotein-associated phospholipase A2 |
| MS4A6A | the same cluster | Fibrinogen, and directly under Alzheimer's disease (a second variant, from the 2019 study) |
| TOMM40 | sits beside APOE | C-reactive protein, cholesterol, cerebral amyloid deposition, and directly under Alzheimer's disease (three variants, from the 2019 study) |
They are not under Alzheimer's, in the rows above, because the position this catalogue happens to hold in each of them was reported by a study asking a different question. A gene is not a disease, and a catalogue records what was measured rather than what a gene is known for.
APP, PSEN1, CR1 and CD33 are not in this catalogue at all, under any heading.
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 162 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.
Willett JDS et al. 2025, Alzheimer's & dementia : the journal of the Alzheimer's Association rs183110257 (BLOC1S3), rs1661880714 (near NSUN4), rs1347265103 (near TBC1D3P2), rs411856 (NECTIN2), rs1368039384 (near GRK5), rs377702 (NECTIN2), rs387369 (NECTIN2), rs1646657652 (PIK3CD), rs1669197221 (LINC02934), rs1179689749 (HOOK3), rs1587397386 (MELK), rs1023546743 (LINC02689) and 63 more — PMID:39998322
Koskeridis F et al. 2024, Nature communications rs35255921 (RELB), rs28615360 (near NECTIN2), rs79638902 (CEACAM16-AS1), rs73566293 (MARK4), rs346758 (MARK4), rs138235833 (near APOC1), rs4605275 (near NECTIN2), rs34995835 (PILRA), rs77399788 (near KRT18P16), rs145538381 (NBEAL1), rs35251323 (EPHA1), rs62119327 (CEACAM19) — PMID:39537608
Kulminski AM et al. 2022, GeroScience rs148734725 (RHOA), rs7614725 (BSN), rs9527702 (near PCDH17), rs72924659 (MS4A6E), rs10808026 (EPHA1), rs55970842 (PSD), rs2305192 (OGA), rs12514965 (near LINC01957), rs141979783 (STAU1), rs2241809 (CHST10), rs61848607 (KCNH1) — PMID:34743297
EADB 2026, Nature genetics rs56965016 (near IGHV3-64), rs8047567 (C16orf95), rs8712 (MAPT), rs376291994 (near RNU6-1325P), rs34450379 (near RNU6-1307P), rs2070217 (CNN2), rs146033453 (VMAC), rs60073596 (TOM1L2), rs144693124 (RBCK1) — PMID:42237039
Jansen IE et al. 2019, Nature genetics rs141864196 (TOMM40), rs10948363 (CD2AP), rs9653111 (CTB-179K24.3), rs183321458 (CTB-179K24.3), rs539159088 (IGSF23), rs11763230 (EPHA1), rs419010 (PVRL2), rs7657553 (HS3ST1), rs4727449 (STAG3) — PMID:30617256
Thomassen JQ et al. 2026, Nature genetics rs75763893 (APH1B), rs17208902 (near TSBP1-AS1), rs113221226 (CASS4), rs9268888 (HLA-DRB9), rs10131116 (DDHD1-DT), rs77892656 (USP3), rs187023552 (NPAS3), rs2274119 (HP1BP3) — PMID:42736378
Sherva R et al. 2020, Alzheimer's & dementia : the journal of the Alzheimer's Association rs79523756 (near KIF12), rs594284 (near LINC01916), rs7985509 (near LINC00564), rs184621768 (near IFITM3P1), rs78049001 (near LINC02039), rs148405458 (LINC02668), rs111535588 (SCN2A) — PMID:32573913
Adewuyi EO et al. 2022, Communications biology rs3774745 (SEMA3F), rs12058296 (PDE4B), rs12561863 (near LINC01720), rs117501883 (KLC3), rs73976310 (ZNF232), rs530324 (SCARA3) — PMID:35851147
Verma A et al. 2024, Science (New York, N.Y.) rs191177039 (SLC6A1), rs186375628 (near NPBWR1), rs574718000 (TBXAS1), rs188208995 (COL5A1) — PMID:39024449
Willett JDS et al. 2026, Alzheimer's & dementia : the journal of the Alzheimer's Association rs374043832 (RGS6), rs889400412 (RALGPS2), rs935208076 (near PNRC2P1) — PMID:41755704
Ge YJ et al. 2024, Alzheimer's & dementia : the journal of the Alzheimer's Association rs10806784 (near RN7SL366P), rs6140401 (near RN7SL547P), rs79872316 (near LINC01851) — PMID:39023044
Sherva R et al. 2023, Molecular psychiatry rs73505251 (ARHGAP45), rs7738720 (near CD2AP-DT), rs73427293 (near TREM2) — PMID:36543923
Herold C et al. 2016, Molecular psychiatry rs188420713 (near LINC01605), rs150511909 (LINC02325) — PMID:26830138
Kunkle BW et al. 2019, Nature genetics rs138190086 (ACE), rs7185636 (IQCK) — PMID:30820047
Li YJ et al. 2023, Alzheimer's & dementia : the journal of the Alzheimer's Association rs9808800 (DSCAM), rs147145199 (ARFGEF2) — PMID:36738287
Bracher-Smith M et al. 2022, Neurobiology of aging rs112437613 (DAB1) — PMID:35977442
Jun G et al. 2016, Molecular psychiatry rs2732703 (KANSL1) — PMID:25778476
Zhang YR et al. 2024, Alzheimer's & dementia : the journal of the Alzheimer's Association rs11554586 (BIN1) — PMID:39129223
Lo MT et al. 2019, Neurobiology of aging rs192775394 (TRAPPC6A) — PMID:30979435
Lambert JC et al. 2013, Nature genetics rs10838725 (CELF1) — PMID:24162737
Sherva R et al. 2025, Alzheimer's research & therapy rs139130053 (near TSPEAR) — PMID:40708016
2026-09-14 · APOE-stratified genome-wide association analyses provide insights into the genetic etiology of Alzheimer's disease. Nature Genetics. 2026. PMID:42736378
Splitting Alzheimer's GWAS by APOE status uncovers 9 genes invisible in the unstratified analysis
Rather than running one genome-wide association study across everyone, this study split Alzheimer's disease GWAS analyses by APOE epsilon-4 and epsilon-2 carrier status -- the single strongest known genetic risk factor for the disease -- and found 9 genomic signals (HP1BP3, SLC50A1, PTPRC, NPAS3, DDHD1, CHST9, SMYD2, PRAMEF1, GFRA1) that only appear once the data is stratified this way, invisible in a standard unstratified analysis. DDHD1 stood out as especially promising: it showed a protective effect specifically in APOE ε4 carriers, was identified as an expression quantitative trait locus, and is separately involved in rare neuronal disease. The motivation is clinical, not just statistical -- recent anti-amyloid drug trials have shown less benefit and more severe side effects in ε4 carriers, and APOE-stratified genetics like this may help explain why and inform which patients specific mechanism-driven therapies should target. This site's own Alzheimer's disease page currently carries 33 variants (APOE, BIN1, ABCA7, CD2AP and others among them) from earlier, unstratified GWAS efforts; none of the 9 genes from this study overlap those. Published 14 September 2026, one day before this entry was written, so it is not yet indexed in GWAS Catalog with rsID-level records this site could independently verify -- this entry documents the study at the condition level rather than adding unverified variant pages.
2026-03-01 · Genetic architecture of plasma pTau217 and related biomarkers in Alzheimer's disease via genome-wide association studies. Alzheimer's & Dementia. 2026. DOI:10.1002/alz.71275
First Korean-population genetics of the blood test now used to help diagnose Alzheimer's
Blood tests for plasma pTau217 have rapidly become a key tool for detecting Alzheimer's disease (AD)-related brain changes without a spinal tap or PET scan, but the genetics behind this biomarker itself -- as opposed to AD diagnosis -- had barely been studied, especially outside European populations. This study ran GWAS on plasma pTau217, pTau181, neurofilament light chain (NfL) and GFAP in 1,972 Korean adults, integrating results with Korean brain single-nucleus transcriptomics and Mendelian randomization. It found genome-wide significant loci for each biomarker: DACT1-DAAM1 for pTau217, KCNJ3 for pTau181, KLHDC4 for NfL, SLC10A7 for GFAP, and PPP4R2 for a composite score. This matters clinically because as pTau217 blood tests move toward routine use for AD screening, understanding genetic influences on the biomarker level itself -- separate from AD risk -- helps interpret results correctly across different populations. This site's Alzheimer's disease page carries 33 variants describing AD risk itself; this is a distinct biomarker-genetics angle, and specifically the first substantial Korean-population data on it.
2026-06-01 · Identification of genetic modifiers of autosomal dominant Alzheimer's disease: a genome-wide association study. The Lancet Neurology. 2026. DOI:10.1016/s1474-4422(26)00123-7
Even a guaranteed-Alzheimer's mutation doesn't fix the age it strikes -- other genes modify that
Autosomal dominant Alzheimer's disease (ADAD) -- caused by mutations in PSEN1, PSEN2 or APP -- essentially guarantees a person will develop Alzheimer's, but the age of onset and clinical course still vary meaningfully between carriers of the same mutation. This study performed whole-genome sequencing on 101 unrelated ADAD mutation carriers from three major cohorts (Knight-ADRC, DIAN, and ADSP) alongside 5,050 asymptomatic controls, searching specifically for genetic modifiers -- genes that don't cause ADAD but shift its timing or severity in people who already carry a causative mutation. This is a mechanistically distinct question from this site's existing Alzheimer's page, which covers risk variants for typical late-onset AD in the general population; ADAD's genetic modifiers instead describe what happens once the disease's occurrence is already certain. Because the underlying cohort is necessarily small (ADAD is rare), specific modifier genes require care in how confidently they are reported, but the study represents real progress toward a therapeutically important question: whether ADAD modifier genes point to intervention targets that could also help delay typical late-onset AD.
2026-07-17 · Nature Communications 2026, PMID:42469217
Verified against the paper's own abstract via Europe PMC (open access, PMID:42469217, published 17 Jul 2026) rather than secondary coverage. Structural, cellular, and genetic study showing a specific mechanism linking two separately-known Alzheimer's risk genes: a splice variant of CD33 (CD33^M), more common in carriers of CD33 risk alleles, binds another Alzheimer's risk protein, clusterin (CLU) -- together with amyloid-beta, but not with APOE -- on the surface of immune cells (monocytes/microglia). That binding suppresses the cells' ability to engulf and clear amyloid plaques. Genetic analysis in the same paper confirms this CD33:CLU interaction tracks with amyloid burden, cognition, and disease risk. Neither CD33 nor CLU/clusterin is currently among the genes on this site's /conditions/alzheimers-disease page (which covers APOE, TREM2, PICALM, BIN1, ABCA7 and others) -- a real, verified gap this news item documents rather than an already-covered finding restated. Filed at the condition level (matches.variants left empty): the paper is mechanistic/functional, not a fresh GWAS association test reporting its own new rsIDs, so there is no specific variant to link yet.
2026-02-01 · Taquet et al., BMC Neurology 2026, PMID:41742112
Shingles vaccine's protection against Alzheimer's is strongest in carriers of one NECTIN2 variant
Verified directly against the paper's own abstract via Europe PMC (PMID:41742112, published 1 Feb 2026). A prospective study of two large cohorts -- the Health and Retirement Study (n=9,188) and UK Biobank (n=66,748), all followed past age 75 -- found that receiving a shingles vaccine between ages 65-75 was associated with significantly lower odds of later Alzheimer's disease (HRS: odds ratio 0.55; UKB: 0.53). The protective association was stronger, and only statistically significant, in carriers of the rs6859 (A) allele in NECTIN2 -- HRS carriers overall: OR 0.43; female HRS carriers specifically: OR 0.28 (a 72% reduction). In non-carriers, the vaccine-AD association was not significant in either cohort. This site's alzheimers-disease page already discusses this same gene (PVRL2 is another name for NECTIN2) via a different variant, rs6857 -- rs6859 is a distinct SNP in the same gene, not the same position. This is an observational association, not a randomized trial of the vaccine for AD prevention, and the authors themselves frame it as supporting further personalized research, not a treatment recommendation.
2026-06-03 · Nature Genetics
Largest Alzheimer's genetics study yet finds 16 new risk regions
A consensus meta-analysis combining every major Alzheimer's GWAS to date -- 128,681 cases (72,721 clinically diagnosed, 55,960 proxy cases from family history) against 849,833 controls -- brought the total number of known Alzheimer's genetic loci to 91, adding 16 never seen before and 25 additional independent risk variants within loci already known. Several of the new loci sit in genes tied to immune signaling (PTPRC/CD45, VAV1, LILRB1/LILRB4, SRC), reinforcing immune dysfunction as a real, recurring thread in the disease's genetics rather than a single-pathway story. PMID:42237039.
2026-06-03 · EADB, et al., Nature Genetics 2026, PMID:42237039
Nine research consortia pooled their European-ancestry studies: 128,681 cases or proxy cases of Alzheimer's disease and related dementias and 849,833 controls. The analysis identified 91 loci, 16 of them new; 56 were detected specifically in clinically diagnosed Alzheimer's disease, and a further 18 are listed as needing outside validation. A polygenic score built from the loci other than APOE went with Alzheimer's pathology rather than other kinds: people in the top tenth of the score had about twice the risk of advanced tangle and plaque pathology at death, compared with people in the middle of the range. 69 of the study's positions are already on this site, 59 of them on the Alzheimer's disease page, including rs429358 (APOE), rs7232 (MS4A6A) and rs17462136 (CASS4). PMID:42237039.
Read this before the list. Fourteen of the variants linked to this page share one trait string, "Alzheimer disease and age of onset", and one source: a family-based analysis of roughly 3,500 people from 1,070 families. They are not that study's headline findings — its genome-wide significant results were at OSBPL6, PTPRG and PDCL3, and none of the fourteen is any of those. They are the tail of one analysis, they are rare, and they should be read as leads rather than as established Alzheimer genes. They are listed because the catalogue records what was published, and hidden results are worse than qualified ones.
Thirty-three more variants, and a different caveat. They come from a 2019 study that reached its large sample size (71,880 cases) by adding "AD-by-proxy" cases — people whose parent had Alzheimer's, not people diagnosed themselves. The study's own justification is a strong genetic correlation between proxy and clinically-diagnosed AD (rg = 0.81), and this page's numbers reflect the combined analysis rather than clinically diagnosed cases alone.
It cannot diagnose. Alzheimer's disease is diagnosed clinically, with cognitive assessment, imaging and increasingly with biomarkers. No variant on this page appears in that process.
It cannot predict. Even APOE, the largest common effect known, shifts a probability. Most people carrying the higher-risk form never develop Alzheimer's disease, and most people who develop it do not carry it.
It cannot rule anything out. The variants that come closest to deciding an outcome — the ones in APP, PSEN1 and PSEN2 — are precisely the ones a common-variant catalogue does not carry. Absence from this page is a fact about the method, not about a person.
Predictive genetic information about dementia is different from most of what is on this site, because there is no treatment that changes what the result means. People who learn a result like this cannot unlearn it, and the decision about whether to look is a real one that clinical genetics services take seriously and handle with counselling before and after.
Nothing on this page is a test result, and it is not a substitute for that conversation. If Alzheimer's disease runs in your family, the useful next step is a doctor or a genetic counsellor.
What a 23andMe/AncestryDNA export or raw VCF can and can't tell you about Alzheimer's Disease comes down to these specific, well-studied positions — not a diagnosis. 563 positions are linked to this page; the ones this page's own text discusses are shown first.
IL6R · rs61812598
See detailed info →PVRL2 · rs6857
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 80 of 563 linked studies with a resolved discovery ancestry.
Databases, guidelines and references
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.
Alzheimer's Disease. MyGeneLog™. https://www.mygenelog.com/conditions/alzheimers-disease
No. Alzheimer's disease is diagnosed clinically, and no variant here appears in that process. Even APOE, the largest common genetic influence known, shifts a probability: most people carrying the higher-risk form never develop the disease, and most people who develop it do not carry it.
Because a common-variant catalogue cannot see them. More than 200 pathogenic variants in those three genes cause early-onset familial Alzheimer's disease, and they are rare changes found by sequencing affected families rather than by scanning for common differences between large groups. Their absence here says something about the method, not about any person.
Because a catalogue records what was measured, not what a gene is famous for. The position this site happens to hold in each of those genes was reported by a study asking a different question. A gene is not a disease, and the same gene turns up under whichever heading somebody happened to study it.
Not all of them. Fourteen come from one family-based analysis of about 3,500 people, and they are not that study's headline findings — its genome-wide significant results were at other genes entirely. They are listed as published leads, labelled as such, because hiding a qualified result is worse than qualifying it.
Talk to a doctor or a genetic counsellor rather than to a website. Predictive information about dementia is unusual in that no treatment changes what a result means, and it cannot be unlearned, which is why clinical genetics services offer counselling before and after testing.
It is a way a 2019 study reached a much larger sample: instead of requiring every participant to have a personal Alzheimer's diagnosis, it also counted people whose parent was diagnosed, identified from UK Biobank family-history questions. That let the study reach 71,880 cases and 383,378 controls. The approach is justified by a strong genetic correlation (0.81) between proxy and clinically diagnosed cases, but it is not the same measurement as a clinical diagnosis, which is why this page states it plainly.
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.