A DNA health report is a set of tables. Each row names a variant (the SNP), the two alleles you carry (your genotype), the allele a study linked to the condition (the risk allele), the size of that link (an odds ratio; a beta in a nutrition report) and the studies behind it. When your file measures enough of a condition's variants, the condition then gets a tendency compared with the average person. This guide reads one real row, then explains the tendency and its limits.
The seven columns
In a Full Human health report, a condition gets a table only when you carry at least one copy of the risk allele at a measured variant, and every such table has the same seven headers. The sample report shows the layout on pages of a complete report.
| Column | What it holds | Example |
|---|---|---|
| SNP | The rsID: a dbSNP reference number for one position in your DNA | rs10995245 |
| Genotype | The two alleles your file reports there, one from each parent | AA |
| Risk Allele | The allele named by the chosen study: among this condition's studies of the variant that give the allele's frequency, the most precisely estimated one, chosen whether or not you carry the allele | A |
| Copies | How many copies of that allele your genotype holds: 1 or 2. Variants with no copy count towards the tendency but are not listed | 2 |
| Frequency | How common the allele is in the study population, as a share of chromosomes between 0 and 1; the average person carries twice this many copies | 0.35 |
| Effect per copy | The odds ratio of one copy, from that most precise study; the health report keeps odds ratios only | OR 1.03 |
| Studies (PubMed IDs) | Every catalogue study of this condition that reports this variant, each linked to its PubMed abstract | 31361310 |
A variant counts for a condition when your file contains its rsID with a call. The call is counted on the DNA strand it fits, judged from the variant's other allele when the catalogue names it (about a quarter of variants) and otherwise from the call's own bases: a call holding the risk allele is read as given; a call holding the risk allele's complementary base together with another base is read on the other strand; a call made only of the complementary base cannot be placed and is not scored; a call with neither base has no copy. The table lists the variants at which you carry one or two copies of the risk allele; variants at which you carry none are not listed, but they still count towards the tendency, below the average person for that variant. These rules are written out on the methodology page.
A worked example: one row of the allergic rhinitis table
Take rs10995245, which the catalogue reports for several traits, in a file whose genotype there is AA.
| SNP | Genotype | Risk Allele | Copies | Frequency | Effect per copy | Studies (PubMed IDs) |
|---|---|---|---|---|---|---|
| rs10995245 | AA | A | 2 | 0.35 | OR 1.03 | 31361310 |
Read it left to right.
- rs10995245 is the position. The same line sits in your raw file; our guide to reading a raw DNA data file shows how to look an rsID up.
- AA means both copies are A, and 2 is what the report counts.
- A is the allele that Johansson et al. 2019 linked to allergic rhinitis, as part of a combined asthma, hay fever and eczema phenotype. With a GG genotype the row would not be in the table, but the variant would still count: no copy of an allele with a frequency of 0.35 is below average.
- 0.35 is the frequency of A in that study's population (0.3471 in the catalogue). The average person carries 0.69 copies, so two copies are 1.31 copies above average.
- OR 1.03 is the odds ratio per copy from the same study (1.035 in the catalogue). The catalogue also reports rs10995245 for narcolepsy (Han et al. 2013, OR 1.23) and for asthma (Ferreira et al. 2019, OR 1.04); those studies stay out of this table, and neither narcolepsy nor asthma has a section in the report at present. An earlier version of the report averaged all three into one effect under allergic rhinitis; it no longer does.
- The PubMed ID is a link in the PDF. It opens the abstract, so you can check the sample size, the population and the exact trait studied.
What an odds ratio of 1.03 means
An odds ratio compares odds, not probabilities. OR 1.03 says that in the study population, the odds of the condition rose by about 3% per copy of the A allele. It does not mean a 3% chance of allergic rhinitis, and it does not add three percentage points to anything.
Turning it into people needs a baseline, which the report does not contain. As an illustration only: if a condition affected 20 people in 100, the odds would be 20 to 80, or 0.25. Multiplied by 1.035 they become 0.259, which is about 20.6 people in 100. One copy with OR 1.03 moves a 20% baseline by about half a person per hundred; on a 1% baseline it adds about one person in three thousand. Most rows sit between 0.9 and 1.3, and the methodology page calls effects below about 1.3 weak.
Beta versus odds ratio
Across our two reports the Effect column holds two kinds of numbers, and they are not interchangeable.
- An odds ratio is used when the outcome is yes or no: a person has the condition or does not. A value of 1.0 means no association; above 1 the allele raises the odds, below 1 it lowers them. Odds ratios are combined by multiplying, which is the same as adding their logarithms; the report adds log odds ratios, each weighted by your copies above or below the average person's.
- A beta is used when the outcome is measured on a scale, such as a blood level. It is the change in that measurement per copy of the allele, in the study's own units, with a sign: positive raises the value, negative lowers it. Betas from different studies come in different units, so the nutrition report counts only the direction of each effect and compares your total with the spread expected around the average.
The health report keeps odds ratios only, so every one of its rows is an odds ratio; betas appear only in the nutrition report, for measured levels such as vitamin D. A beta of 0.05 for a blood level says nothing about an odds ratio of 1.05 for a disease.
How the tendency is computed
Each condition has an Analysis paragraph above its table: a sentence with the coverage (how many variants the catalogue knows, how many your file measures, how many could be compared with the average person) and a sentence with the tendency, which also fills the summary table at the front of the PDF with one of Higher odds, Lower odds, No meaningful difference, Not comparable or Not assessable. The method is simple on purpose, and the methodology page publishes it in full.
- One study per variant. Among the catalogue studies of this condition for a variant that give the allele's population frequency (directly or through another study of the same allele), the most precisely estimated one (the narrowest confidence interval) gives the risk allele, the effect per copy and the allele's population frequency, whether or not you carry that allele. Nothing is averaged.
- Compared with the average person. Each measured variant counts with your copies of the risk allele minus the copies the average person carries, which is twice the frequency (once on a man's X chromosome when the file shows one base). Two copies of an allele with a population frequency of 0.5 are one copy more than average; no copy of it is one copy fewer.
- Combination and limits. Those centred copies are multiplied by each variant's log odds ratio and added; the sum, taken back to an odds ratio, is your odds relative to the average person: 1.20 or more prints as higher odds, 0.83 or less as lower odds, anything between as no meaningful difference. A person with an average genotype at every variant lands at 1.00, however many variants the condition has.
- Coverage. A tendency is only given when your file measures at least a third of the independent variants the catalogue knows for the condition, and at least a third of those known variants can be compared with the average person (each needs a population frequency, an effect on the condition's scale and a call that fits the strand); otherwise the condition prints as not assessable (fewer than a third of the known variants measured) or not comparable (fewer than a third of the known variants compared), and the variants you carry are still listed.
What the method leaves out, by design:
- One lead variant per region of 250,000 bases is kept per condition, but finer linkage between variants is not modelled.
- No absolute baseline risk enters, so higher odds says nothing about absolute probability.
- One study per variant is taken rather than a meta-analysis, and studies are not weighted by sample size.
- The average person is the study population's, usually of European ancestry; the comparison transfers imperfectly to other ancestries.
- Strand is judged from the variant's other allele only when the catalogue names it, about a quarter of variants; for the rest it is judged from the call's own bases, so a call made only of the risk allele's complementary base is not scored, and an A/T or C/G variant reported on the opposite strand can be misread.
This is a transparent aggregation of published effects, not a clinical polygenic risk score. A tendency can change between catalogue releases as studies are added or re-curated; the release in use (2026-06-01 at the time of writing) is stated on the methodology page.
Why a "Higher odds" line is not a diagnosis
Three things stand between a tendency and your health.
Baseline risk is missing. An odds ratio only scales a starting point the report does not know. A Higher odds tendency for a rare condition can describe a probability that is still very small.
The cohorts are mostly European. The GWAS Diversity Monitor counted 88.4% of genome-wide association study participants as European in its data check of 6 October 2026 (gwasdiversitymonitor.com, checked 14 October 2026). Effects measured in one ancestry transfer imperfectly to others, and the report applies them unchanged.
Your chip covers a fraction of the catalogue. A consumer array reads a few hundred thousand positions, often around a quarter of the catalogue's lead variants, so a condition gets a tendency only when your file measures at least a third of its known variants, and that tendency rests on those variants alone. Two files from two providers can give different reports for the same person.
Behind all three: an association is a statistical signal in a study population, not a cause. Where the signals come from and how they are curated is the subject of our deep dive into the GWAS Catalog.
The 17 categories
The catalogue's EFO and MONDO ontology mapping groups conditions into 17 categories; categories are not sold as separate reports. A condition is printed only when the catalogue has rows for it and the report has a text for it (Alzheimer's disease, asthma, schizophrenia, COVID-19 and narcolepsy have catalogue rows but no text yet), and a category with no printed condition is skipped, so a complete report has 14 sections for a file whose sex is male or undetermined and 16 for a female file, with up to about 190 conditions. The section names below are the ones printed in the PDF, so you can match them to its contents page; the examples are conditions the report prints.
- Autoimmune and Inflammatory Diseases: psoriasis, rheumatoid arthritis, Crohn's disease
- Cancer: breast carcinoma, lung carcinoma, melanoma
- Cardiovascular Diseases: atrial fibrillation, coronary artery disease, hypertension
- Congenital: coeliac disease
- Dental and Oral Health: dental caries, periodontitis
- Dermatological: acne, skin sensitivity to sun, chemotherapy-induced alopecia
- Endocrine and Metabolic Disorders: gout, obesity, hypothyroidism
- Gastrointestinal: irritable bowel syndrome, gastroesophageal reflux disease (GERD), pancreatitis
- Genetic and Metabolic: cystic fibrosis associated meconium ileus, the category's only condition, which has no rows in the current catalogue release, so the section is not printed at present
- Infectious: hepatitis B, malaria, leprosy
- Musculoskeletal: osteoporosis, carpal tunnel syndrome, adolescent idiopathic scoliosis
- Neurological and Mental Health Disorders: major depressive disorder, epilepsy, age-related macular degeneration
- Obstetrics and Gynecology: pre-eclampsia (the section's only condition at present; female files only)
- Other: tooth agenesis, exfoliation syndrome
- Reproductive: endometriosis, uterine fibroid (female files only)
- Respiratory: allergic rhinitis, chronic obstructive pulmonary disease, obstructive sleep apnea
- Urological: nephrolithiasis (kidney stones; the section's only condition at present)
Some sections are short because few of their conditions have genome-wide significant associations with a reported effect size. Pharmacogenomics, carrier status, rare pathogenic variants such as those in BRCA1 and BRCA2, ancestry, and candidate-gene variants such as the common MTHFR ones are not interpreted (MTHFR rs1801133 appears only in the nutrition report, as a gene highlight next to homocysteine, not as a score).
What to bring to your doctor
A report like this is a prompt for a conversation, not a result in itself. Bring:
- The PDF, opened at the summary table, so the tendencies can be read at a glance.
- The rows you want to ask about, with their PubMed IDs. A doctor or genetic counsellor can read the study behind a row and say whether it applies to someone of your age, sex and ancestry.
- Your family history. A pattern in close relatives carries more weight than any row in the table.
- What the report is not. It is not a clinical genetic test; if one is warranted, a doctor can order it.
Not generated a report yet? Start with your raw data file; our guide to downloading your DNA data from 23andMe, AncestryDNA and MyHeritage covers each provider.
FAQ
What is a risk allele?
At a given position, people carry one of two or more variants, called alleles. The risk allele is the allele whose effect the study reports, and the catalogue keeps it whichever way that effect points. Despite the name, its odds ratio can be below 1, which means carrying it was associated with lower odds of the condition.
Why is a SNP missing from my report?
Five reasons cover almost every missing row: your chip did not read that position, or read it as a no-call; you carry no copy of the risk allele, so the row is not shown; the study did not state a single-base risk allele; the association failed the catalogue filters for significance, effect size and confidence interval; or a more strongly associated variant within 250,000 bases was kept for the condition instead.
Why do different tools disagree about the same SNP?
Tools differ in sources, filters and arithmetic. One reads SNPedia summaries, another the GWAS Catalog with a significance threshold; a third weights copies of an allele or corrects for correlated variants. Two honest tools can print different results from the same file. Compare the study each one cites before trusting either.
What do "Not assessable", "Not comparable" and "No meaningful difference" mean?
Not assessable: your file measures fewer than a third of the variants the catalogue knows for the condition, so the report gives no tendency; the variants you do carry are still listed. Not comparable: enough variants are measured, but fewer than a third of the known variants can be compared with the average person (each needs a population frequency, an effect on the condition's scale and a call that fits the strand), so the report gives no tendency. No meaningful difference: enough variants are measured and compared, and your odds relative to the average person stay between 0.83 and 1.20. None of the three means your risk is average in a clinical sense; the report compares you with the average person of the study population and knows no baseline risk.
This article is for general information only and is not medical advice. Talk to a qualified healthcare professional before making health decisions based on genetic data.
Got your raw file?
Turn it into your health report now.
Open the export in your browser; Full Human reads it locally and writes a PDF with the study behind every row. Nothing is uploaded.
409 catalog traits · 17 categories · GWAS Catalog June 2026
Your DNA file stays on your device during report generation.