Ves al contingut principal

Aquest contingut encara no està disponible en català. El mostrem en anglès.

Equip DNA Info Lab · Revisió editorial · · 10 min de lectura

Pharmacogenomics from Raw DNA Data: What Your Genes Say About Medications

Can a raw DNA file say anything about medications? Yes, within limits. Consumer files from 23andMe, AncestryDNA and MyHeritage can contain many of the gene variants that pharmacogenomic guidelines (CPIC, curated in PharmGKB) use to describe drug response — for clopidogrel, codeine, simvastatin, warfarin and others; which ones a given file holds depends on the provider and the chip version. What an array cannot detect are structural variants like CYP2D6 gene duplications. The annotations are research context for a conversation with your doctor — never a reason to change a dose on your own.

The same pill, at the same dose, can be ineffective in one person and cause side effects in another. A large part of that difference is written in the genes that metabolise drugs — and a surprising amount of it is already sitting in the raw DNA file you can download from 23andMe or AncestryDNA.

What pharmacogenomics is

Pharmacogenomics (PGx) studies how genetic variation changes the way a person responds to a drug: how fast it is activated or broken down, how strongly it acts, and how likely side effects are. The field has matured from anecdotes into published, peer-reviewed prescribing guidance — the CPIC guidelines (Clinical Pharmacogenetics Implementation Consortium) and the PharmGKB database, which curates gene–drug annotations with graded levels of evidence: level 1A means variant-specific prescribing guidance exists in a clinical guideline or an FDA-approved drug label, while level 3 marks a low level of evidence.

The central actors are enzymes with memorable names: CYP2D6, CYP2C19, CYP2C9, plus transporters like SLCO1B1 and targets like VKORC1. Variants in these genes can make you break a drug down unusually slowly or unusually fast — the difference between a standard dose doing nothing and the same dose becoming an overdose.

What a raw DNA file can actually tell you

Consumer arrays do not sequence your genes; they measure pre-selected positions. Many of the variants the CPIC guidelines rely on are single-letter variants (SNPs), and several have been measured on consumer chips: a 2017 analysis of the pharmacogenetic reports 23andMe offered in the UK lists every variant below, and 23andMe's current test still reports three CYP2C19 variants and one SLCO1B1 variant. Which positions your own file contains depends on the provider and the chip version; where they are present, the file already holds the relevant genotype:

  • CYP2C19*2 (rs4244285, a no-function allele) and CYP2C19*17 (rs12248560, an increased-function allele) — the alleles behind CPIC's guidance for clopidogrel and for several antidepressants; both are simple SNPs.
  • SLCO1B1 rs4149056 (c.521T>C) — the simvastatin myopathy variant, a single SNP.
  • VKORC1 rs9923231 and the CYP2C9 *2 and *3 alleles (rs1799853, rs1057910) — central inputs to the warfarin dosing algorithms, all SNPs.

When you upload your file to our report, the pharmacogenomics section matches your genotypes against PharmGKB annotations and shows each match with the drug, the gene, its PharmGKB level of evidence and the published annotation behind it — research context, in plain language, with the source linked.

The honest limit: SNPs yes, star alleles not always

Here is the caveat most consumer pages skip. Pharmacogenetic star alleles are not all single-letter variants. Some are structural: whole-gene deletions or duplications. The most important case is CYP2D6 — an enzyme responsible for the oxidative metabolism of up to a quarter of commonly prescribed drugs (Samer et al., 2013):

  • An ultrarapid CYP2D6 metaboliser often carries extra copies of the gene (CPIC writes these duplications as "xN", 2021 guideline). An array measuring individual positions cannot count copies, so it cannot see this.
  • A whole-gene deletion (*5) is invisible to SNP positions for the same reason.
  • The result: an array can genotype the SNP-based CYP2D6 alleles it covers, but it cannot certify a complete CYP2D6 phenotype. "No variant found" on an array is not proof of a normal metaboliser; a 2017 analysis notes that CYP2D6 cannot be assessed fully with a genome-wide genotyping chip, because of gene deletions and its similarity with pseudogenes.

Any tool that prints a definitive CYP2D6 metaboliser status from consumer array data is overstating what the file contains. Ours reports the per-genotype annotations for the variants actually measured, and says so.

Four worked examples

CYP2C19 and clopidogrel

Clopidogrel is a prodrug: CYP2C19 must activate it. Intermediate metabolisers (one no-function allele, such as *2) and poor metabolisers (two of them) form less of the active drug, and after coronary stenting they show a higher risk of major adverse cardiovascular events and stent thrombosis. That is why the 2022 CPIC guideline recommends avoiding clopidogrel in both groups in acute coronary syndrome or stenting, and using an alternative such as prasugrel or ticagrelor if there is no contraindication (PMID 35034351). The key alleles are SNPs that a raw file can contain.

CYP2D6 and codeine

Codeine is also a prodrug — CYP2D6 converts it to morphine. Poor metabolisers form very little morphine and get diminished pain relief; ultrarapid metabolisers form more, with a higher risk of toxicity. The 2021 CPIC guideline recommends avoiding codeine in both groups (PMID 33387367), and notes that several regulators advise against codeine in children under 12 and that the US FDA label advises against breastfeeding while taking it. The SNP alleles can be on the chips, but the ultrarapid case is usually a duplication — exactly the structural variant an array cannot detect. This example is the honest limit in action.

SLCO1B1 and simvastatin

SLCO1B1 transports simvastatin into the liver. The rs4149056 C allele (c.521T>C) decreases that transport, which increases exposure to simvastatin and the risk of muscle damage (myopathy). The 2022 CPIC statin guideline, which replaced the 2012 one, recommends an alternative statin when the SLCO1B1 genotype predicts decreased function, or a limited simvastatin dose if simvastatin is still warranted (PMID 35152405). One SNP, measured directly.

VKORC1, CYP2C9 and warfarin

Warfarin dosing is notoriously individual. VKORC1 rs9923231 changes how sensitive the drug target is, and the CYP2C9 *2 and *3 alleles change how fast the drug is cleared; together with clinical factors they drive the genotype-guided dosing in the 2017 CPIC guideline, which also uses CYP4F2 and rs12777823 and gives recommendations specific to continental ancestry (PMID 28198005). All three variants discussed here are SNPs.

What "poor" and "rapid" metaboliser mean

PGx guidelines classify people into metaboliser phenotypes per enzyme. For CYP2C19, the CPIC guideline uses five: poor, intermediate, normal, rapid and ultrarapid; CYP2D6 has poor, intermediate, normal and ultrarapid. For a drug that is active as taken, a poor metaboliser clears it slowly — it accumulates, and standard doses can overshoot — while a fast metaboliser clears it quickly and may get less effect. For a prodrug the direction flips: a poor metaboliser makes little of the active drug (clopidogrel, codeine), and an ultrarapid CYP2D6 metaboliser makes more morphine from codeine. The label is per enzyme, not per person: you can be a normal CYP2C19 metaboliser and an intermediate CYP2D6 one at the same time.

What to do with this information

  1. Do not change any medication on your own. Not the dose, not the schedule, not the drug. PGx annotations describe group-level associations, not your prescription.
  2. Bring it to your doctor or pharmacist. A PharmGKB annotation plus your genotype is a legitimate starting point for that conversation — CPIC guidelines are written to help clinicians act on genotype results that are already available.
  3. If a real prescribing decision depends on it, get a clinical-grade test. Hospital-grade PGx panels include the copy-number analysis arrays lack, and are interpreted under clinical quality standards.

Frequently asked questions

Can my 23andMe raw file tell me how I will respond to a drug?▼

Partially. It can contain the SNPs behind many guideline annotations — clopidogrel, simvastatin, warfarin and others — so it can flag published gene–drug associations that apply to your genotypes. What it cannot do is certify a complete enzyme phenotype, because structural variants like CYP2D6 duplications are invisible to arrays.

What is the difference between PharmGKB and CPIC?▼

PharmGKB is the curated database: it collects gene–drug annotations and grades the evidence behind each. CPIC is the consortium that turns the strongest of that evidence into clinical prescribing guidelines. PharmGKB is where the annotations live; CPIC is what clinicians act on.

If my file shows no risky variant, does that mean any dose is safe?▼

No. An array only measures the positions on the chip. It cannot see CYP2D6 copy-number changes, rare variants, or anything about your liver, kidneys, other medications or diet — all of which affect drug response. Absence of a finding is not evidence of safety.

Why can't the array see CYP2D6 duplications?▼

A microarray measures which letters you carry at fixed positions — it does not count how many copies of a gene you have. Detecting duplications or deletions requires copy-number analysis, which is a different laboratory method used by clinical PGx panels.

Should I get a clinical pharmacogenetic test instead?▼

If a specific prescribing decision depends on it — for example before certain antiplatelet, antidepressant or oncology treatments — yes: ask your doctor about a clinical-grade panel. For general curiosity and for informing that conversation, your existing raw file is a reasonable and much cheaper start.

Is any of this medical advice?▼

No. It is research context: published associations between genotypes and drug response, with sources linked. Dosing decisions belong to your doctor and pharmacist, who see the whole clinical picture.

Related reading

See your pharmacogenomic annotations

Upload your 23andMe, AncestryDNA or MyHeritage file (zipped or extracted) and get the PharmGKB annotations that match your genotypes, with each source linked. Free coverage summary; full report 10 €. Private, no kit needed.

Related articles

Aquest article té finalitats exclusivament educatives i no constitueix consell mèdic, diagnòstic ni tractament. Consulta sempre un professional sanitari per a decisions sobre la teva salut.