· 6 min read · Reviewed by the DNA Info Lab editorial team
23andMe Raw Data Analysis: What You Can (and Cannot) Learn
What can a 23andMe raw data analysis tell you? Polygenic risk scores, pharmacogenomics (drug response), trait associations and some carrier-status hints — all from common variants. It cannot give a diagnosis, detect rare mutations not on the chip, or predict your future. Here is the honest scope.
Once you have your raw file, the useful question is not "what can I run" but "what can the result actually mean". Consumer microarray data is powerful for some questions and useless for others. Here is the honest boundary.
What you CAN learn
- Polygenic risk scores — your relative risk for common conditions (type 2 diabetes, cardiovascular, Alzheimer's, depression and more), summed across hundreds of common variants from the GWAS Catalog.
- Pharmacogenomics — how your genotype may affect response to common drugs (warfarin, statins, clopidogrel, codeine, SSRIs, metformin), from PharmGKB.
- Trait associations — well-studied, low-stakes traits like lactose tolerance, caffeine metabolism, or the ACTN3 "sprint" variant.
- Carrier hints — some ClinVar-flagged variants that are on the chip. Not a full carrier screen, but a starting signal.
What you CANNOT learn
- A diagnosis. A risk score is a population statistic, not a verdict about you. High risk ≠ disease; low risk ≠ safety.
- Rare private mutations. If a variant was not on the chip (~99.98% of your genome isn't), it simply is not in the file. A "clear" result never rules out a rare mutation.
- Your full genome. Microarrays read ~650,000 chosen positions. For everything else you need whole-genome sequencing.
- The future. Genetics is one input among many — lifestyle, environment and chance dominate most outcomes.
The ancestry caveat
Roughly 90% of the participants in the studies behind these scores are of European ancestry. If your ancestry is different, the risk multipliers may not transfer with the same magnitude — an honest analysis flags this rather than hiding it. Treat non-European results with extra caution.
How accurate is it?
The genotyping itself is highly accurate — over 99% concordance with sequencing at the positions the chip measures. The uncertainty is in interpretation: effect sizes come from population studies with confidence intervals, and coverage varies (a category matched by 40 variants is more reliable than one matched by 4). Good tools show you both the score and its confidence.
How to run the analysis
You can do it by hand — look up individual rsids in SNPedia and the GWAS Catalog — or upload the whole file to a tool that cross-references it automatically and explains each result, with its limits, in plain language.
Frequently asked questions
Can 23andMe raw data diagnose a disease?▼
No. It produces relative risk estimates and trait associations, not a clinical diagnosis. Anything concerning should be confirmed with a doctor and, where appropriate, clinical-grade testing.
Will an analysis find every genetic condition I might carry?▼
No. It only covers variants on the chip (~0.02% of the genome). A reassuring result never rules out a rare mutation that was not tested.
Is the analysis accurate for non-European ancestry?▼
Less so. Most underlying studies over-represent European-ancestry participants, so risk estimates are weaker for other ancestries. Interpret with caution.
Do I need whole-genome sequencing instead?▼
Only if you need rare-variant coverage. For common-variant risk, pharmacogenomics and traits, microarray raw data is more than enough.
Run your analysis
Upload your 23andMe, AncestryDNA or MyHeritage file for a free report — 11 trait categories, pharmacogenomics and ClinVar flags, each with its confidence and sources.
Related articles
This article is for educational purposes only and does not constitute medical advice, diagnosis or treatment. Always consult a healthcare professional for decisions about your health.