Pharmacogenomics in Raw DNA: What Your File Can Tell You

CYP2C19, CYP2C9, VKORC1, SLCO1B1 and TPMT variants sit in most raw files. What they mean for common drugs, what SNP chips miss, and why not to self-adjust.

Pharmacogenomics - how genetic variation changes the way people respond to medicines - is the part of consumer genetics with the strongest evidence and the most direct clinical use. It is also the part where a raw DNA file is most likely to mislead you if you read it without knowing what the chip does and does not capture. Here is what is genuinely in your file, what it means, and where the edges are.

Why drug genes matter

Most medicines are cleared by a small set of liver enzymes, and several of those enzymes come in versions that work faster or slower than average. A slow version can leave a drug lingering at higher levels; a fast version can clear it before it acts. For a drug that has to be activated by the enzyme, the logic flips. The Clinical Pharmacogenetics Implementation Consortium (CPIC) publishes guidelines that translate these genotypes into prescribing advice, and regulators including the FDA maintain a table of drug-gene associations with established evidence.

The variants most raw files contain

The examples below are common, well-studied, and readable from a single SNP each. Alleles are given on the forward strand, as raw files report them.

CYP2C19 and clopidogrel. Clopidogrel, a blood thinner given after stents and some strokes, must be activated by CYP2C19. The 2 loss-of-function allele is rs4244285 (variant allele A);17, a gain-of-function allele, is rs12248560 (variant T). People with two loss-of-function alleles activate the drug poorly, and CPIC recommends an alternative for them. CYP2C19 also affects several antidepressants and proton-pump inhibitors, usually in the direction of dose adjustment rather than avoidance.

CYP2C9 and VKORC1 with warfarin. Warfarin dosing varies several-fold between people, and about a third of that variation is genetic. CYP2C9 2 (rs1799853, variant T) and3 (rs1057910, variant C) slow its breakdown; the VKORC1 promoter variant rs9923231 (variant T) makes the drug’s target more sensitive. Carriers of these need lower doses, and published algorithms combine genotype with age, weight and other drugs to estimate a starting dose.

SLCO1B1 and statins. rs4149056 (variant C) reduces the transporter that moves simvastatin into the liver, raising blood levels and the risk of muscle side effects. CPIC’s guidance is to prefer a lower dose or a different statin in carriers, not to avoid statins.

TPMT and thiopurines. Azathioprine and mercaptopurine, used in autoimmune disease and leukaemia, can cause dangerous bone-marrow suppression in people with low TPMT activity. rs1142345 (variant C) marks the 3C allele;3A also involves rs1800460. This is one of the oldest examples of genotype-guided dosing in routine care.

Genespiral’s health analysis reads several of these as an educational “medication response” category, and each marker’s page describes the associated genotypes.

What the chip cannot see

This is the part that turns an interesting file into a hazardous one if ignored.

  • CYP2D6 is largely unreadable. It metabolises a quarter of common drugs, including codeine, tramadol, many antidepressants and antipsychotics, and tamoxifen. Its important variation includes whole-gene deletions and duplications and dozens of star alleles defined by combinations of variants. A SNP chip reads a few of them and cannot count gene copies at all. Any CYP2D6 call derived from a raw file is incomplete.
  • HLA alleles are tagged, not typed. The severe skin reactions to abacavir (HLA-B57:01) and carbamazepine (HLA-B15:02) are predicted by HLA typing. Some chips carry a tag SNP that tracks the allele well in some populations and poorly in others.
  • Rare variants are absent. Chips read common variation. A rare loss-of-function variant that a clinical sequencing panel would catch is simply not on the chip.
  • Phasing is unknown. If you carry two different variants in one gene, the file does not say whether they sit on the same copy or opposite copies, which can change the interpretation.

Clinical pharmacogenomic tests are designed around these gaps: they type the full star alleles, count CYP2D6 copies, and report a phenotype (poor, intermediate, normal, rapid, ultrarapid metaboliser) rather than a list of SNPs.

How to use this well

The right response to finding a pharmacogenomic variant in your file is small and specific: keep a note of it, and mention it to a prescriber when one of the relevant drugs comes up. A clinician can then decide whether to order a proper test or simply take it into account. The wrong responses are to change a dose yourself, stop a medicine, or refuse one because of a chip result. Drug choice depends on the whole picture, and a raw-file genotype is one incomplete input.

If you are new to reading genotypes, our explainer on heterozygous and homozygous results covers the basics, and why the same variant can look different between files explains the strand issue that trips up drug-gene lookups more than any other.

References

This article is educational only and is not medical advice. Never change a prescribed medicine or its dose based on a consumer DNA file.

Further reading