One of the most heritable common conditions there is — and one of the least predictable from a genome. Twin studies put the inherited share around 70-80%, yet the first reliable common variants were only found in 2019, and each one moves the odds by a few percent. This page is about why both of those are true at the same time.
ADHD affects about 5% of children and 2.5% of adults. It runs in families about as strongly as height does. And there is no genetic test for it, nor is one close.
Those three sentences are not in tension once you see the shape of the trait, and that is what this page is for.
Twin and family studies have consistently put the heritability of ADHD at roughly 70-80% — near the top of the range for any common condition. That number describes how much of the variation between people in a population tracks with their genetic similarity. It says the genes matter enormously.
What it does not say is that a small number of genes matter. Decades of candidate-gene work produced almost nothing that replicated. The first genome-wide significant results for ADHD arrived in 2019, from a study of 20,183 diagnosed people and 35,191 controls, and it found 12 independent positions.
Twelve positions, each shifting the odds by a few percent per copy, in a trait that is 70-80% heritable. The inherited component is real and it is spread across thousands of positions, almost all of them too small to see even at that sample size.
Beyond the twelve positions, the paper made three points that matter more than any single rsID.
The signals were enriched in constrained parts of the genome — regions that have changed little across evolution, and genes that do not tolerate being broken. Whatever these variants are doing, they are doing it in parts of the genome that matter.
They were enriched around brain-expressed regulatory marks. Most of them are not in genes at all; they are in the sequence that decides how loudly nearby genes are read, in brain tissue.
And the genetic signal from diagnosed ADHD matched the signal from ADHD symptoms measured across the general population. That is the finding with the most weight outside genetics: it supports the view that a diagnosis is the far end of a continuous trait everybody sits somewhere on, not a separate category of person.
All six come from that 2019 meta-analysis. Two sit in genes with a plausible story — SORCS3, a brain sorting receptor, and PTPRF, involved in building connections between neurons. One, near DUSP6, sits beside a gene tied to dopamine signalling, which is the system ADHD medicines act on. The other three are in sequence with no gene attached — which is what most genome-wide findings look like, and an honest thing to show rather than hide.
A seventh, listed in the catalog against FOXP2, is not resolvable in the current Ensembl release, so it is not here. We would rather have five positions we can check than six we cannot.
It cannot diagnose. ADHD is identified by a clinician from behaviour, observed in more than one setting, over time, with other explanations ruled out. No genotype enters that assessment anywhere.
It cannot predict. The strongest position here changes the odds by about 8% per copy, against a background risk of a few percent. Nothing in that is visible in one person.
It cannot choose a medicine. Stimulant response is not predicted by these positions, and there is no pharmacogenomic guideline for ADHD medicines built on them.
Because "your genes explain most of this, and we still cannot read them" is a real and useful thing to understand — and because the alternative, on the open internet, is a page that promises far more.
It is also the clearest available answer to a bad argument: that a highly heritable condition must be a fixed one. ADHD is as heritable as almost anything in this catalogue, and it is also strongly shaped by sleep, structure, schooling and support. Both of those are true, and the genetics is the reason neither one wins.
A separate 2019/2020 study asked a different question: not which positions raise ADHD risk on their own, but how much of ADHD's genetics overlaps with a different trait, lifetime cannabis use. Using genome-wide summary statistics from an ADHD study of 53,293 people and a cannabis-use study of 32,330 people, it estimated the genetic correlation between the two at r2 = 0.29 — a real, moderate overlap, not the same trait twice.
The same paper ran a two-sample Mendelian randomization analysis and found support for ADHD being causal for later cannabis use: people with ADHD had roughly 8 times the odds (OR 7.9) of using cannabis compared with people without it. That is a genetics-based causal estimate, not a claim about any individual, and it lines up with what clinicians already treat as a known risk: ADHD is associated with a higher rate of substance use generally.
This site holds four variants from that study's records — rs113551349 (SLC6A9), rs797267 (POC1B), rs56319043 (near KDM4A-AS1) and rs2782639 (PTPRF — the same gene as rs3001723 above, from a different study of ADHD alone). None of these four is one of the paper's own headline findings. Its own two new genome-wide-significant single-variant loci from the cross-trait analysis, rs145108385 and rs4259397, are not in this catalogue at all — this page's four variants are other associations from the same GWAS Catalog record, not the featured result, and are presented as that rather than as something they are not.
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 24 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.
van der Laan CM et al. 2025, Nature genetics rs4749820 (near RNA5SP299), rs7708883 (near ASS1P10), rs2025284 (COL19A1), rs13405307 (near TMEM18), rs71428363 (ABCA12), rs7100410 (STK32C), rs1435746 (SEMA6D), rs4131786 (CDH8), rs9959583 (ZNF521), rs4462678 (DCC) — PMID:40962958
Demontis D et al. 2019, Nature genetics rs56135409 (SGO1-AS1), rs1592757 (—), rs2243517 (RNF219-AS1), rs4275621 (—) — PMID:30478444
Demontis D et al. 2023, Nature genetics rs2311059 (IQCF1), rs114142727 (VGLL3), rs6537401 (LSM6), rs76857496 (near TMEM200C) — PMID:36702997
Yao X et al. 2021, Translational psychiatry rs12265655 (SORCS3), rs10789368 (LINC01360) — PMID:33479212
Verma A et al. 2024, Science (New York, N.Y.) rs554999570 (near TRAK1) — PMID:39024449
LaBianca S et al. 2024, Nature genetics rs12819667 (near DUSP6) — PMID:38036780
Soler Artigas M et al. 2020, Molecular psychiatry rs139868495 (LINC01288) — PMID:30610198
Martin J et al. 2018, Biological psychiatry rs8039398 (SEMA6D) — PMID:29325848
2025-09-17 · Genome-wide association meta-analysis of childhood ADHD symptoms and diagnosis identifies new loci and potential effector genes. Nature Genetics. 2025. DOI:10.1038/s41588-025-02295-y
This study combined two different ways of measuring ADHD: continuous symptom scores from 290,134 measurements across 70,953 people (multiple raters, ages, instruments) and a separate binary ADHD diagnosis GWAS. The symptom-score analysis alone found no genome-wide significant variants, but combining it with the diagnosis data pushed the total to 39 independent loci, 17 of them newly reported -- the strongest signal being rs549845 (P=3x10^-19). Using a fine-mapping method designed to identify likely causal effector genes rather than just nearest-gene guesses, the study nominated 22 candidate ADHD genes implicating new biological pathways not previously linked to the condition, though this site could not independently verify specific gene names beyond the study's own reporting. It found moderate negative genetic correlations with cognitive traits, and -- most practically -- polygenic scores built from the combined symptom-plus-diagnosis data outperformed scores built from either measure alone across three validation cohorts. The core conceptual finding: this supports treating clinical ADHD as the extreme end of a continuous trait indexed by symptom severity in the general population, rather than a wholly separate biological category, and shows that including subclinical symptom data (not just formal diagnoses) helps find genes standard case-control designs miss. This site's ADHD page carries 7 variants; the 17 newly reported loci are not currently among them.
Architecture. ADHD heritability from twin studies is approximately 70-80%. The 2019 meta-analysis (20,183 cases, 35,191 controls) identified variants surpassing genome-wide significance at 12 independent loci — the first robust common-variant associations reported for the disorder. Per-allele odds ratios for the loci recorded here are in the range of approximately 1.03 to 1.08. Associations were enriched in evolutionarily constrained genomic regions, in loss-of-function-intolerant genes, and around brain-expressed regulatory marks, consistent with a highly polygenic architecture acting largely through regulation rather than coding change.
Replication and continuity with population traits. Three replication samples were examined: a clinically diagnosed cohort, a self-reported ADHD sample, and a meta-analysis of quantitative measures of ADHD symptoms in the general population. All supported the primary findings, while showing study-specific differences in genetic overlap with educational attainment. The strong concordance between the diagnosed-case GWAS and quantitative population symptom measures supports the interpretation that clinical ADHD represents an extreme expression of continuous heritable traits.
The variants recorded here. rs11591402 (SORCS3, P = 4e-9), rs1592757 (5q21, P = 1e-9), rs4275621 (11p14, P = 3e-9), rs1427829 (near DUSP6, P = 7e-9), rs3001723 (PTPRF, P = 1e-8) and rs2243517 (RNF219-AS1, P = 4e-8). Positions, alleles and 1000 Genomes frequencies were verified individually against Ensembl GRCh38. The catalog's FOXP2 entry, rs5886709, does not resolve in the current Ensembl release and is excluded.
Clinical boundary. ADHD is a clinical diagnosis based on developmental history and observed impairment across settings, per DSM-5 and ICD-11 criteria. No genotype is diagnostic, contributory or exclusionary. Polygenic scores for ADHD have no established clinical role and are not validated for individual prediction. There is no pharmacogenomic guideline covering stimulant selection or dosing on the basis of these positions; CYP2D6 status is relevant to atomoxetine dosing and is a separate matter from the risk loci described here.
What a 23andMe/AncestryDNA export or raw VCF can and can't tell you about ADHD (Attention Deficit Hyperactivity Disorder) comes down to these specific, well-studied positions — not a diagnosis. 31 positions are linked to this page; the ones this page's own text discusses are shown first.
SLC6A9 · rs113551349
See detailed info →POC1B · rs797267
See detailed info →PTPRF · rs2782639
See detailed info →KDM4A-AS1 · rs56319043
See detailed info →near TRAK1 · rs554999570
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 13 of 31 linked studies with a resolved discovery ancestry.
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 |
|---|---|---|
| CYP2D6 | Atomoxetine | Atomoxetine is a non-stimulant used for ADHD, and CYP2D6 polymorphisms influence its metabolism and thereby both its efficacy and its safety. CPIC published dosing recommendations by CYP2D6 genotype in 2019. This is about how much drug a person is exposed to at a given dose; it is not a prediction of whether ADHD treatment will help, and no genotype is used to decide whether to treat. (CPIC Guideline for CYP2D6 Genotype and Atomoxetine Therapy (Clin Pharmacol Ther 2019, PMID 30801677).) |
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.
ADHD (Attention Deficit Hyperactivity Disorder). MyGeneLog™. https://www.mygenelog.com/conditions/adhd
Largely, yes: twin studies put the heritability of ADHD at roughly 70-80%. But it is not passed on by a single gene. The inheritance is spread across thousands of common variants, each shifting the odds by a few percent, so no genotype can say whether a child will have ADHD.
No, and there is not one on the horizon. ADHD is highly heritable but the inheritance is spread across thousands of positions with tiny individual effects. The first reliable ones were found in 2019, and the strongest of them moves the odds by about 8% per copy. Diagnosis is clinical.
Heritability describes how much of the variation across a population tracks with genetic similarity. It says nothing about how few genes are involved. For ADHD the answer is: thousands, each too small to see. Height works the same way and is far easier to measure.
No — and ADHD is the clearest counter-example in this catalogue. It is about as heritable as anything here, and it also responds to sleep, structure, schooling, support and medication. Heritable and changeable are not opposites.
No. None of these positions predicts stimulant response, and no guideline uses them for prescribing. CYP2D6 status matters for atomoxetine dosing, but that is a separate question from the risk positions on this page.
Because most genome-wide findings are not in genes. They fall in the regulatory sequence between them — DNA that changes how loudly a nearby gene is read. Working out which gene, and in which tissue, is a separate piece of research that has not been done for every one of these.
A 2019/2020 genetic study found real overlap between ADHD and lifetime cannabis use (genetic correlation r^2=0.29) and, using a Mendelian-randomization method, estimated that ADHD is causal for later cannabis use (about 8x the odds). That is a population-level genetic estimate from one study, not a prediction about any individual, and clinicians already treat elevated substance-use risk as a known part of ADHD.
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.