Single-variant pharmacogenomic findings from 23andMe v5 raw data · 100% Han Chinese ancestry · Build GRCh37/hg19 · 594,952 usable biallelic autosomal SNPs after QC
This page replaces an earlier April 2026 report that performed a polygenic-risk-score (PRS) cross-reference against psychiatric GWAS using EUR-cohort summary statistics. That analysis didn't apply: the subject is 100% Chinese ancestry, and EUR-cohort GWAS weights don't transfer cleanly to East-Asian individuals (different LD patterns, different allele frequencies, sometimes flipped effect direction). The PRS magnitudes from that report were uncalibrated single-subject sums driven mostly by chr 6 MHC tag-SNP homozygosity — not interpretable as personal risk. This rewrite drops the PRS framing and reports only what's interpretable from a single 23andMe v5 chip without forced statistics: well-characterized single-variant pharmacogenomic findings with literature-backed functional reads. The methodological argument behind the rebuild is documented in a companion paper: Ancestry-matched GWAS removes the chromosome 6 MHC dominance in consumer-chip psychiatric polygenic risk scoring (PDF, 12 pages), with two interactive companions: /genomics/eas-scz-landscape/ visualizes the empty MHC band in the Lam 2019 EAS-SCZ GWAS, and /genomics/chip-prs-explorer/ lets you drop your own 23andMe v5 file (browser-only, no upload) to see how many of your typed SNPs land in the MHC band.
This is for research / educational purposes only. Not a clinical pharmacogenomic test. Several clinically important East Asian variants — CYP2D6 *10 (rs1065852), ADH1B *2 (rs1229984), the HLA-B alleles tagged for carbamazepine and allopurinol risk — aren't covered by 23andMe v5. A proper CPIC-aligned pharmacogenomic panel would close those gaps. Don't make medical decisions from this page.
Headline pharmacogenomic profile. East-Asian (EAS) baseline frequencies in parentheses for context — they describe how typical the genotype is for someone of Han Chinese ancestry, not how rare or important the finding is.
| Drug class / area | Gene · variant | Genotype | Functional read | EAS context |
|---|---|---|---|---|
| PPIs · clopidogrel · SSRIs | CYP2C19 *2 (rs4244285) CYP2C19 *3 (rs4986893) |
GGGG |
*1/*1 — extensive (normal) metabolizer | ~40% of EAS (60% are PM/IM via *2 or *3) |
| Warfarin (vitamin-K antagonist) | VKORC1 −1639 (rs9923231) CYP2C9 *2 (rs1799853) CYP2C9 *3 (rs1057910) |
CTCCAA |
Intermediate warfarin sensitivity via VKORC1 het; CYP2C9 *1/*1 normal | EAS-typical is VKORC1 hom for low-dose allele (~80%); intermediate (CT) is the more EUR-like ~20% |
| Statins (simvastatin) | SLCO1B1 *5 (rs4149056) | TT |
*1A/*1A — normal transport, no elevated myopathy risk | EAS *5 freq ~13% |
| 5-Fluorouracil / capecitabine | DPYD *2A (rs3918290) | CC |
*1/*1 — normal DPD activity | Typical (EAS *2A <1%) |
| Efavirenz / methadone / bupropion | CYP2B6 *6 (rs3745274) | GG |
*1/*1 — normal metabolism | Typical (EAS *6 ~25%) |
| Irinotecan · bilirubin handling | UGT1A1 *6 (rs4148323) | GG |
No EAS-specific Gilbert variant from this position | EAS *6 freq ~15-20%; you're hom reference |
| Tacrolimus / calcineurin inhibitors | CYP3A5 *3 (rs776746) | CC (strand-flip ambiguity) |
Likely *3/*3 nonexpresser (typical EAS, lower tacrolimus dose required) — but verify on a clinical panel before acting | ~75% of EAS are nonexpressers |
| HCV peginterferon-α + ribavirin | IFNL3 / IL28B (rs12979860) | CC |
Favorable HCV treatment response genotype | ~90% of EAS — typical |
| Alcohol | ALDH2 *504Lys (rs671) | AG |
Glu/Lys heterozygote — mild flush, slower acetaldehyde clearance, mildly elevated risk for upper-aerodigestive carcinogen exposure if drinking | ~40% of Chinese — typical |
| Caffeine | CYP1A2 *1F (rs762551) ADORA2A (rs5751876) |
AACC |
Fast caffeine metabolizer (half-life ~3-4h) + low caffeine-induced anxiety | Combination is interpretable cross-ancestry |
| Late-onset Alzheimer's risk modifier | APOE (rs429358 + rs7412) | TT + CC |
ε3/ε3 — most common APOE form globally; neither AD-risk (ε4) nor AD-protective (ε2) | Cross-ancestry interpretable |
CYP2C19 is one of the most clinically actionable pharmacogenes. Variants *2 (rs4244285) and *3 (rs4986893) are loss-of-function; *3 is essentially East-Asian-specific.
Genotype: *1/*1 (extensive / normal metabolizer)
About 60% of East Asians carry at least one loss-of-function CYP2C19 allele — substantially more than Europeans (~30%). Being *1/*1 puts the subject in the ~40% of East Asians with normal CYP2C19 metabolism.
Warfarin dose is one of the best-validated pharmacogenomic dosing equations. The IWPC (2009) algorithm uses VKORC1 (target) + CYP2C9 (clearance) + clinical factors to predict therapeutic dose.
| Variant | Genotype | Functional | EAS allele freq |
|---|---|---|---|
| VKORC1 −1639G>A (rs9923231) | CT | Intermediate warfarin sensitivity | A allele ~90% in EAS (vs ~40% in EUR) |
| CYP2C9 *2 (rs1799853) | CC | No *2 (typical EAS) | EAS *2 <1%, EUR ~12% |
| CYP2C9 *3 (rs1057910) | AA | No *3 | EAS *3 ~5%, EUR ~7% |
Most East Asians are VKORC1 A/A homozygotes and require lower warfarin doses than Europeans (~3 mg/day vs ~5+ mg/day). This subject is heterozygous, predicting an intermediate dose requirement — between the EAS-typical and EUR-typical patterns. CYP2C9 is wild-type *1/*1, so warfarin clearance is normal.
Strand note: 23andMe v5 reports rs9923231 on the dbSNP forward strand (C/T alleles); the warfarin literature describes the same variant as −1639G>A on the gene-coding strand. Chip C = literature G (high-dose / EUR-typical allele); chip T = literature A (low-dose / EAS-typical allele). Subject's CT = literature GA = heterozygous.
If warfarin were ever prescribed, the IWPC dose-prediction model is appropriate to use; the predicted dose for this VKORC1 + CYP2C9 combination is in the typical-EAS range, neither at the very-low-dose tail nor the EUR-typical higher end. Direct oral anticoagulants (dabigatran, apixaban, rivaroxaban) bypass both pathways and are not affected by this profile.
SLCO1B1 *5 (rs4149056 T>C) reduces hepatic statin uptake, leading to higher systemic exposure and increased risk of statin-induced myopathy — best characterized for simvastatin, where the SEARCH trial showed *5/*5 carriers have ~7-fold increased myopathy risk.
Genotype: SLCO1B1 *1A/*1A (rs4149056 TT) — normal transport function
No elevated risk of statin-induced myopathy from this position. CPIC recommends standard simvastatin dosing for *1A/*1A. This is a genuinely useful negative finding — many patients are unaware they have an elevated myopathy risk variant.
| Gene · variant | Genotype | Drugs affected | Implication |
|---|---|---|---|
| DPYD *2A (rs3918290) | CC (*1/*1) |
5-fluorouracil, capecitabine | Normal DPD activity; standard chemo dosing. |
| CYP2B6 *6 (rs3745274) | GG (*1/*1) |
Efavirenz, methadone, bupropion, ketamine | Normal metabolism; standard dosing. |
| UGT1A1 *6 (rs4148323) | GG |
Irinotecan, atazanavir; bilirubin handling | No EAS-specific Gilbert-pattern reduction from this allele. UGT1A1 *28 (the EUR-common Gilbert variant, a TA repeat polymorphism) is not directly typed by 23andMe v5. |
| CYP3A5 *3 (rs776746) | CC strand-flip note |
Tacrolimus, cyclosporin, sirolimus | Likely *3/*3 nonexpresser → standard tacrolimus dose works fine (the typical EAS pattern; ~75% are nonexpressers). Some 23andMe v5 SNPs are reported on the reverse strand, so verify on a clinical panel before acting on this. |
| IFNL3 / IL28B (rs12979860) | CC |
Peginterferon-α + ribavirin (HCV) | Favorable response genotype. C allele freq ~90% in EAS (~80% CC homozygous) vs ~70% in EUR (~35% CC homozygous). Less practically relevant in the era of direct-acting HCV antivirals, but historically the strongest predictor of treatment response. |
Two genes dominate East-Asian alcohol pharmacology: ADH1B (ethanol → acetaldehyde) and ALDH2 (acetaldehyde → acetate). The combination determines flush severity, drinking tolerance, and upper-aerodigestive carcinogen exposure.
| Variant | Genotype | EAS context |
|---|---|---|
| ALDH2 *504Lys (rs671) | AG — Glu/Lys heterozygote |
~40% of Chinese — typical. Mild flush. ~10% of Chinese are Lys/Lys homozygous (severe flush, can drink almost zero alcohol). |
| ADH1B *2 (rs1229984) | not typed by 23andMe v5 | ~70% of EAS carry at least one *2 allele — fast ethanol clearance. Without this measurement we can't predict whether the flush will be mild-mild (slow ADH1B + slow ALDH2 het) or pronounced (fast ADH1B + slow ALDH2 het). |
The ALDH2 het + smoking combination is a well-established carcinogen-exposure risk: acetaldehyde is a Group 1 carcinogen (IARC 2009), and ALDH2 *504Lys heterozygotes who drink and smoke have substantially elevated esophageal squamous-cell carcinoma risk (Brooks et al. 2009 PLoS Med). The mitigation is straightforward: limit alcohol, don't smoke, accept the flush as a biological warning.
Disulfiram (Antabuse) and similar ALDH inhibitors should not be used in ALDH2 *504Lys carriers — the flush reaction is already partially present and amplification can be dangerous.
Two independent variants jointly determine caffeine response:
The combination — *1F/*1F CYP1A2 + low ADORA2A anxiety sensitivity — is a clean "caffeine well-tolerated" pharmacogenomic profile, especially under inducer-state conditions.
Single-variant findings at well-characterized neurotransmitter loci. Effect sizes for individual variants are typically small; these are descriptive, not diagnostic.
| Variant | Genotype | EAS context | Functional read |
|---|---|---|---|
| COMT Val158Met (rs4680) | AG — Val/Met |
~50% globally | Heterozygous; intermediate catecholamine clearance. Population-mean phenotype on the warrior/worrier continuum. |
| BDNF Val66Met (rs6265) | CT — Val/Met |
Met allele is more common in EAS (~40%) than EUR (~20%); heterozygote is the modal EAS genotype | One Met copy. Mildly impaired activity-dependent BDNF secretion; effect sizes on memory/anxiety are small in well-powered studies. |
| DRD2 Taq1A (rs1800497) | AG — A1/A2 |
A1 allele freq ~40-50% in EAS (vs ~25-30% EUR); A1 carrier is typical for Chinese | ~15-20% reduced striatal D2 binding compared to A2/A2. |
| OPRM1 Asn40Asp (rs1799971) | GG — Asp/Asp |
Asp40 freq ~30-40% in EAS (vs ~10-15% EUR); by HWE ~10-15% of EAS are Asp/Asp homozygous | Reduced μ-opioid receptor signaling. Best-replicated clinical translation: better naltrexone response in alcohol use disorder (PharmGKB Level 2A clinical annotation; CPIC's 2021 opioid guideline addresses select opioids but not naltrexone specifically). |
| OXTR (rs53576) | AA |
A allele freq ~50% in EAS (vs ~30% in EUR); AA homozygote is ~25% of EAS but only ~10% of EUR | The "lower social sensitivity" associations in this SNP came from EUR studies where AA was the minor-allele homozygote (~10%); in EAS the A and G alleles are roughly balanced, so the EUR-derived "minor allele = trait" framing doesn't generalize. The biological interpretation is therefore weaker than commonly cited. |
| HTR2A (rs6313) | AA |
A allele common in both EUR and EAS | Modulates 5-HT2A signaling. Not significant in well-powered psychiatric GWAS; descriptive only. |
APOE is the single largest common-variant genetic risk modifier for late-onset Alzheimer's disease. Three alleles (ε2, ε3, ε4) are defined by the combination of two SNPs:
| SNP | Genotype |
|---|---|
| rs429358 | TT |
| rs7412 | CC |
APOE ε3/ε3 — the most common form globally (~60% of populations).
Neither the AD-risk-elevating ε4 allele (rs429358-C) nor the AD-protective ε2 allele (rs7412-T) is present. ε3/ε3 is the population-baseline genotype — neither risk-elevating nor risk-reducing for late-onset Alzheimer's. Cross-ancestry interpretable.
23andMe v5's ~600K SNP chip wasn't designed as a pharmacogenomic panel and misses several variants that are clinically actionable for East Asian patients. The biggest gaps:
| Gene · variant | What it tells you | EAS frequency |
|---|---|---|
| ADH1B *2 (rs1229984) | Speed of ethanol → acetaldehyde conversion. Combined with ALDH2 status, determines flush severity and alcohol dependence risk profile. | ~70% of EAS carry *2 (vs ~5% EUR) |
| CYP2D6 *10 (rs1065852) | Reduced CYP2D6 activity. Affects codeine activation, tramadol, atomoxetine, many tricyclic antidepressants, antipsychotics. CPIC has detailed dosing guidelines for *10/*10 patients. | *10 allele freq ~40-50% in EAS → ~70-85% are *10 carriers (≥1 copy); ~10-25% are *10/*10. EUR *10 allele freq ~5-10%, carriers ~10-20%. |
| HLA-B*15:02 (multi-SNP haplotype) | FDA black-box warning: ~10-fold increased risk of carbamazepine-induced Stevens-Johnson syndrome / toxic epidermal necrolysis. FDA recommends testing in patients of EAS ancestry before carbamazepine prescription. | ~5-15% in Han Chinese, Thai, Malay, Vietnamese; rare in EUR/AFR |
| HLA-B*58:01 (multi-SNP haplotype) | Allopurinol severe cutaneous adverse reaction (SCAR) risk. CPIC recommends an alternative urate-lowering agent in *58:01 carriers. | ~5-10% in Han Chinese, ~15% in some Southeast Asian populations |
| CYP2C19 multi-allele coverage | v5 covers *2 and *3 (the major loss-of-function alleles in EAS). Minor alleles like *17 (gain-of-function, increased activity) are absent. For the EAS PM/IM determination this is mostly fine, but a clinical panel would be more comprehensive. | — |
For anyone seriously contemplating these medications (especially carbamazepine, allopurinol, or codeine), a CPIC-aligned clinical pharmacogenomic panel — not a consumer chip — is the appropriate test.
23andMe v5 mostly reports SNPs on the dbSNP forward strand, but for some positions (notably some CYP3A5 and CYP2D6 variants) the reported alleles are on the reverse strand. Where a result is strand-ambiguous, this page flags it. CPIC-aligned clinical panels resolve strand explicitly and are the appropriate source for clinical decisions.
scripts/build_pharmacogenomics.py — reads the 23andMe TSV and matches against a curated table of well-characterized pharmacogenomic positions.