Learn Guide
A medication report from 23andMe, AncestryDNA or MyHeritage raw data: what your file can answer
A raw data file from 23andMe, AncestryDNA or MyHeritage can be turned into a medication report in a few minutes. How much that report can say depends on which positions your particular chip typed, and that changes with the company and with the year. Across ten real exports we measured, the medication positions our report could read ran from 12 of 14 down to 1.
Key takeaways
- From a chip export our report can read seven medication genes in full: ABCG2, CYP2C19, CYP2C9, CYP4F2, NUDT15, SLCO1B1 and VKORC1. Every drug their cards quote is part of a CPIC level A gene and drug pair[9].
- Exports from the same company differ by year: an AncestryDNA file from 2018 let the report read 7 of its 14 medication positions and one from 2024 read 12. The 2018 MyHeritage export we measured read 3.
- Twelve of the fourteen is the most any chip can read. The other two are A/T changes a chip cannot orient, so no chip reads DPYD or HLA-B in full.
- CYP2D6 needs aligned reads. In UK Biobank, CYP2D6 groups inferred from imputed chip data matched the combined calls for 64.9 percent of people[2].
- The FDA's position is that a consumer pharmacogenetic result should be confirmed with clinical testing before any medical decision, and that it should not be used to stop or change any medicine[11].
A medication report reads genes that change how the body handles particular drugs: liver enzymes that activate or clear them, transporters that move them in and out of cells, and the protein warfarin acts on. The Clinical Pharmacogenetics Implementation Consortium, CPIC, grades each gene and drug pair from A to D. At level A, genetic information should be used to change prescribing of the drug, and only level A and B pairs carry a prescribing recommendation[1]. Variation in these genes is common: among 487,409 UK Biobank participants, 99.5 percent had genotypes in 14 CPIC genes that predict an atypical response to at least one drug[2].
Seven genes a chip can read in full
Most genes in a medication report rest on a few known positions, and a genotyping chip measures fixed positions. When a chip measured every position a gene's result depends on, it can answer that gene. Our report reads nine medication genes from fixed positions, and seven of them can be read in full from a chip export. The table lists them with the positions each card reads and the drugs its CPIC guidance covers.
| Gene | What it does | Positions the card reads | Drugs on the card |
|---|---|---|---|
| ABCG2 | A pump that pushes rosuvastatin back out of gut and liver cells | rs2231142 (Q141K) | Rosuvastatin[3] |
| CYP2C19 | A liver enzyme that activates clopidogrel and clears several antidepressants | *2, *3 and *17 | Clopidogrel[4]; escitalopram and citalopram[5] |
| CYP2C9 | A liver enzyme that clears warfarin, several NSAIDs and fluvastatin | *2 and *3 | Celecoxib, ibuprofen and other NSAIDs[6]; warfarin[7]; fluvastatin[3] |
| CYP4F2 | An enzyme in vitamin K recycling | *3 (rs2108622) | Warfarin[7] |
| NUDT15 | An enzyme that breaks down thiopurine drugs | *3 (rs116855232) | Azathioprine and mercaptopurine[8] |
| SLCO1B1 | A liver transporter that pulls statins out of the blood | *5 (rs4149056) | Atorvastatin, fluvastatin, lovastatin, pitavastatin, pravastatin, rosuvastatin and simvastatin[3] |
| VKORC1 | The protein warfarin acts on | rs9923231 (-1639G>A) | Warfarin[7] |
Source: The report's chip cards and the CPIC guidance they quote. Every gene and drug pair in the table is CPIC level A in CPIC's database export of 13 June 2026[9].
Where a drug's guidance depends on two genes, the card quotes the row for the pair of results the file gave. Rosuvastatin is the case on a chip: CPIC's statin guideline gives its starting dose from SLCO1B1 and ABCG2 together, so the ABCG2 card reads its line with your SLCO1B1 result beside it[3].
Two more genes in the same part of the report can never be read in full from a chip. DPYD and HLA-B each have a defining position where the change swaps A for T. A chip reports the two letters it saw without saying which strand of the DNA it read, and for an A/T change both strands give the same two letters, so the reading cannot be oriented. A whole-genome VCF reads both genes.
CYP2D6, the gene behind CPIC's codeine, tramadol and tamoxifen guidance, is a different problem. Its variation includes deleted, duplicated and hybrid copies of the whole gene, and a chip has no way to count copies. In about 50,000 UK Biobank participants with both chip and exome data, CYP2D6 groups inferred from the imputed chip data matched the study's combined calls for 64.9 percent of people, against 99.4 percent for CYP2C19[2]. Our report calls CYP2D6 only from aligned reads, and the guide to star alleles shows why counting copies decides the result.
What ten real exports let the report read
We ran the report's own chip reader over ten public chip exports, the files listed on our data page[10]. Fourteen positions feed the medication cards: ten for the seven genes above and four for DPYD and HLA-B, two of which no chip can orient. The table counts how many of the fourteen each file let the report read, and which of CYP2C19's three positions it read.
| Chip export | Medication positions read, of 14 | CYP2C19 positions read |
|---|---|---|
| 23andMe v3, 2014 | 11 | *2, *3, *17 |
| 23andMe v4, 2020 | 11 | *2, *3, *17 |
| 23andMe v5, 2019 | 12 | *2, *3, *17 |
| AncestryDNA v2, 2018 | 7 | *3 |
| AncestryDNA v2, 2024 | 12 | *2, *3, *17 |
| MyHeritage, 2018 | 3 | *2, *3 |
| FamilyTreeDNA, 2020 | 9 | *3 |
| FamilyTreeDNA, 2023 | 3 | *2, *3 |
| Living DNA, 2019 | 1 | None |
| Genes for Good, 2025 | 8 | *2, *3 |
Source: Aimosti measurements of 6 October 2026 (all fourteen positions) and 9 October 2026 (CYP2C19), using the report's chip reader on public Personal Genome Project exports[10]. A position counts as read when the file gives a genotype the report can place on the right strand.
12 of 14
medication positions read by the best exports, 23andMe v5 and AncestryDNA 2024[10]
1 of 14
read by the weakest, a Living DNA export from 2019[10]
4 of 10
exports that let the report read all three CYP2C19 positions[10]
Within one company the year can matter as much as the name. AncestryDNA's 2018 export let the report read 7 of the 14 positions and its 2024 export 12, while FamilyTreeDNA's went the other way, from 9 in 2020 to 3 in 2023. The three 23andMe exports, from chip versions v3, v4 and v5, read 11 or 12. We had one MyHeritage export, from 2018, and it read 3, two of them CYP2C19's *2 and *3. Each row is one person's file, so a file of yours from another year may read more or less, and the file check tells you which.
CYP2C19 shows what one missing position costs. Six of the ten exports did not read *17, the allele that raises the enzyme's activity. Without it, a file that carries neither *2 nor *3 could belong to a normal, a rapid or an ultrarapid metabolizer, so the report leaves the group open[4].
What the card says when a position is missing
For the seven genes above, a chip card has four outcomes, set by what the file read. The quoted words are the card's own.
| What the file read | What the card shows |
|---|---|
| Every defining position | The result and CPIC's guidance for it, with the note that "this call is read directly, not assumed" |
| Some positions, no variant among them | No phenotype, the metabolizer or function group that drug guidance depends on. The card names the positions it could not read and calls the gap "not examined", never a reference result |
| Some positions, a variant among them | The variant it found, with no phenotype and no dosing line "because the call could not be completed, not because nothing was found", and a note to take what was read to the prescriber |
| A variant on each copy of the gene | The full result, because a missing position "cannot change a diplotype whose two copies are both already assigned" |
Source: The chip branch of the report's medication module, as of 12 November 2026.
How to judge any medication report from a raw data file
Several services take a chip export and return a medication report. These questions work for any of them, ours included.
- Does it say, gene by gene, which defining positions your file contained and which it lacked?
- What does it do with a position the file lacks? A report that fills the gap with the common version will print a normal result for a gene it never saw.
- Does it name the guideline behind each drug line, and its level? CPIC makes prescribing recommendations only for level A and B pairs[1].
- Does it give a CYP2D6 result from a chip? In UK Biobank, CYP2D6 groups inferred from chip data disagreed with the combined calls for about a third of people[2].
- Does it say where the file is processed and stored, and whether it is deleted afterwards?
- Does it tell you to confirm a result before acting on it? For the consumer pharmacogenetic report it authorised in 2018, the FDA required a warning that the results should not be used to stop or change any medication[11].
What our report shows, and what it costs
The chip report reads the seven genes above wherever your file allows, and each card quotes CPIC for the result it holds. The same report covers traits, maternal and paternal lines, ancestry, and polygenic scores given as a band. It costs €19 in Europe and $19 elsewhere, paid once, with no subscription. The file is processed and stored in Finland and deleted after analysis by default.
Before you pay, the file check reads your export inside your browser, uploads nothing, and lists by name the medication genes your file can answer. A sequenced genome reads more: from a plain VCF the report calls nine medication genes, from a gVCF 25, and from a BAM or CRAM the Deep Read adds CYP2D6 with its copy number. The star alleles guide has the full table by file type.
Before you act on a result
CPIC's levels are about prescribing: at level A, genetic information should be used to change prescribing of the drug[1]. Prescribing is your clinician's decision, made with information a DNA file does not hold. When the FDA authorised a consumer pharmacogenetic report in 2018, the agency put its position this way.
Consumers should not use this test to make treatment decisions on their own. Any medical decisions should be made only after discussing the results with a licensed health care provider and results have been confirmed using clinical pharmacogenetic testing.
A raw data report is best used as a reason to ask. The people who can act on it are a prescriber or a pharmacist, when a medicine on the list comes up, and in the FDA's position a clinical test confirms the result before any dose changes because of it[11].
What Aimosti would (and wouldn't) show you
From a chip export the report reads up to seven medication genes: ABCG2, CYP2C19, CYP2C9, CYP4F2, NUDT15, SLCO1B1 and VKORC1. A gene gets a result only when the file read every one of its defining positions, and the card then quotes CPIC's guidance for that result. A gene the file read in part gets no metabolizer or function group, and its card names what was read and what was missing. The file check lists by name which of these genes your file reads before you pay.
What we won't claim
We won't call CYP2D6 from a chip, read a missing position as the common version of a gene, or tell anyone to start, stop or change a medicine. A pharmacogenomic result is a fact about one gene for a prescriber to weigh, and the FDA's position on consumer pharmacogenetic reports is that results should be confirmed with clinical testing before any medical decision.
Bottom line. A raw data file can answer part of a medication report, and the size of that part depends on the chip it came from. The best exports we measured let the report read 12 of its 14 medication positions; the weakest read one. A report worth having says, gene by gene, which positions it had, and leaves a gene open when it did not.
Questions people ask
Can I get a medication report from my AncestryDNA raw data?
Yes, for the genes your file's chip read. Of the two AncestryDNA exports we measured, the 2018 file let our report read 7 of its 14 medication positions and the 2024 file 12, all three CYP2C19 positions among them. The file check lists which genes your own file reads before you pay. CYP2D6 cannot be read from any chip.
Does 23andMe raw data include pharmacogenomic variants?
Yes. The three 23andMe exports we measured, from chip versions v3, v4 and v5, let our report read 11 or 12 of its 14 medication positions and all three CYP2C19 positions. The v5 export read all 12 positions a chip can read, which covers the seven chip-readable genes in full.
Is MyHeritage raw data enough for a medication report?
It depends on the file. The one MyHeritage export we measured, from 2018, let our report read 3 of the 14 medication positions, CYP2C19 *2 and *3 among them but not *17, so most genes stayed open. A newer file may read more. Run the file check on yours before you buy.
Can raw data from a consumer DNA test show my CYP2D6 status?
Not reliably. CYP2D6 varies by whole copies of the gene, which a chip cannot count, and in UK Biobank chip-based CYP2D6 groups matched the combined calls for 64.9 percent of people[2]. Our report reads CYP2D6 only from aligned reads, a BAM or CRAM file.
Which medicines can a raw data medication report cover?
In our report, CPIC's guidance for seven statins, warfarin, clopidogrel, escitalopram and citalopram, azathioprine and mercaptopurine, and NSAIDs such as celecoxib and ibuprofen, each when the file read the gene in full. Every one of these is a CPIC level A pair with its gene[9].
Should I change my medication based on a raw data report?
No. A change of medicine is a prescriber's decision, and the FDA's position on consumer pharmacogenetic reports is that results should be confirmed with clinical pharmacogenetic testing before any medical decision[11].
References
- Assignment of CPIC levels for genes/drugs. Clinical Pharmacogenetics Implementation Consortium (CPIC), via ClinPGx. Read 2026-11-12 from the page's data endpoint. Level A: genetic information should be used to change prescribing of affected drug. Only level A and B gene/drug pairs have sufficient evidence for at least one prescribing action to be recommended.
- McInnes G, Lavertu A, Sangkuhl K, Klein TE, Whirl-Carrillo M, Altman RB. Pharmacogenetics at scale: an analysis of the UK Biobank. Clinical Pharmacology & Therapeutics, 2021. doi:10.1002/cpt.2122 Abstract: 487,409 participants; for 14 CPIC pharmacogenes, 99.5% may have an atypical response to at least 1 drug. Table 1, imputed chip data against the integrated call set: CYP2D6 phenotype 64.86%; CYP2C19 phenotype 99.44%.
- Cooper-DeHoff RM, Niemi M, Ramsey LB, et al. The Clinical Pharmacogenetics Implementation Consortium guideline for SLCO1B1, ABCG2, and CYP2C9 genotypes and statin-associated musculoskeletal symptoms. Clinical Pharmacology & Therapeutics, 2022. doi:10.1002/cpt.2557
- Lee CR, Luzum JA, Sangkuhl K, et al. Clinical Pharmacogenetics Implementation Consortium guideline for CYP2C19 genotype and clopidogrel therapy: 2022 update. Clinical Pharmacology & Therapeutics, 2022. doi:10.1002/cpt.2526
- Bousman CA, Stevenson JM, Ramsey LB, et al. Clinical Pharmacogenetics Implementation Consortium (CPIC) guideline for CYP2D6, CYP2C19, CYP2B6, SLC6A4, and HTR2A genotypes and serotonin reuptake inhibitor antidepressants. Clinical Pharmacology & Therapeutics, 2023. doi:10.1002/cpt.2903
- Theken KN, Lee CR, Gong L, et al. Clinical Pharmacogenetics Implementation Consortium guideline (CPIC) for CYP2C9 and nonsteroidal anti-inflammatory drugs. Clinical Pharmacology & Therapeutics, 2020. doi:10.1002/cpt.1830
- Johnson JA, Caudle KE, Gong L, et al. Clinical Pharmacogenetics Implementation Consortium (CPIC) guideline for pharmacogenetics-guided warfarin dosing: 2017 update. Clinical Pharmacology & Therapeutics, 2017. doi:10.1002/cpt.668
- Relling MV, Schwab M, Whirl-Carrillo M, et al. Clinical Pharmacogenetics Implementation Consortium guideline for thiopurine dosing based on TPMT and NUDT15 genotypes: 2018 update. Clinical Pharmacology & Therapeutics, 2019. doi:10.1002/cpt.1304
- CPIC gene-drug pairs and levels. Clinical Pharmacogenetics Implementation Consortium (CPIC), via ClinPGx, 2026. Read from CPIC's database export of 13 June 2026 (api.cpicpgx.org/v1, pair table): every gene and drug pair in the seven-gene table is level A.
- What your DNA file can actually read: measured on 14 real files. Aimosti, 2026. Lists the ten public chip exports measured here, by vendor and year.
- FDA authorizes first direct-to-consumer test for detecting genetic variants that may be associated with medication metabolism. U.S. Food and Drug Administration, press announcement, 2018. Read 2026-11-12. Quotes Tim Stenzel; the special controls include a labeling requirement for a warning that the consumer should not use the test results to stop or change any medication.
Last reviewed . Every number on this page links to the source it comes from; if one of them has moved, tell us.
Free · in your browser
Find out what your own file can read.
The free check opens your DNA file on your device and shows which format you have and what a report could read from it. The file is never uploaded.