1. Pharmacogenomics — how you process drugs

(PharmCAT 3.2.0 on WGS; verified.)

No-Call ≠ "normal" — and these are now all recovered. In the position-limited PGx VCF, CYP2D6, CYP2C9, DPYD, TPMT, NUDT15, UGT1A1, NAT2 returned No Call and SLCO1B1 was ambiguous (~50 candidate diplotypes) — because those positions were missing from the VCF, not because the data was absent. All have since been re-genotyped directly from the 34× BAM (§6): every one resolved, all normal-function / standard-risk. The lesson generalizes — absence of a finding in a position-limited VCF means not measured, not reassuring.

2. Disease-risk variants (ClinVar / ACMG)

• No high-penetrance dominant pathogenic variants found (0 high-confidence ACMG-actionable). Reassuring.
• Hemochromatosis / hereditary iron overload — genetically excluded. All three classic HFE variants are wild-type (C282Y G/G, H63D C/C, S65C A/A; 18–25×, NCBI-verified positions), so the disease genotype can't exist; and you carry none of 490 ClinVar pathogenic/likely-pathogenic variants across the rare non-HFE iron genes (HJV, HAMP, TFR2, SLC40A1, + CP/FTL; all covered 35–36×, so genuine absence). Fits EAS ancestry — HFE C282Y is a Northern-European founder variant. (Genetics ≠ iron status; clinical iron overload, if ever flagged by ferritin/transferrin-sat, is a separate lab question.) Detail: output/specialist/iron/results.md.
• Recessive carrier states (single allele — relevant for family planning, not your own health):
• MPZL2 — carrier, recessive hearing loss (2-star Pathogenic; the most solid of the three).
• MUTYH — carrier, MUTYH-associated polyposis (recessive; 1-star). Monoallelic carriers have at most a small colorectal-risk bump; aligns with standard screening.
• FOXL2 — 1-star Likely-pathogenic; its conditions (BPES / premature ovarian insufficiency) are female-specific → not relevant to you.
• The other 130 "exploratory" P/LP hits were 1-star Conflicting interpretations = background noise.

3. Structural variants (SV / CNV)

• No pathogenic (class-5) calls. Six class-4 (likely-pathogenic) small deletions flagged by AnnotSV: SLC25A24, ADAMTS17, PLIN4, PRKRA (×2 / CHROMR), DSPP (136 bp – 6.4 kb), plus a class-3 VPS13B duplication.
• These hit mostly recessive genes (SLC25A24, ADAMTS17, PRKRA→DYT16, VPS13B→Cohen) or repeat-rich regions (PLIN4, DSPP) where short-read callers over-call. Zygosity is not recorded in this table. Short-read SV calls are false-positive-prone → none should be given clinical weight without orthogonal confirmation (clinical array / MLPA / long-read), and only if a matching phenotype exists.

4. Polygenic risk (ancestry-calibrated to East Asian)

• Trustworthy signal: Stroke ⚠ and Type-2-diabetes ⚠ elevated; CAD and AFib average (the scary "100th percentile" cardiac flags were a platform artifact — proven by a height null-control). Lower-than-average: Alzheimer's, depression, hypertension. Higher-than-average: asthma, COPD, lung cancer, lipids, gout, RA.
• Read the magnitudes as "elevated / not elevated," never as exact percentiles — European genome-wide PRS are systematically inflated in this setup. Full analysis + caveats: output/pgsc_calc/percentile_results.md.

Trait PRS — height & cognition (interest only, not predictive for you)

You asked to see these. They are European-GWAS scores with near-zero individual predictive validity for you — curiosities, not facts. Two reasons they don't transfer: (a) you're EAS 100% and these were built in European cohorts (cross-ancestry portability is poor — worst for education/cognition, where much of even the within-population signal is population stratification + assortative-mating + "genetic-nurture"/environmental confounding, not direct effects on the brain); (b) this platform's documented inflation (the height null-control, below).

The kicker: six education/cognition scores — spanning 2018–2024, up to 6.7M variants, and including the 3-million-person EA4 flagship — scatter from the 17th to the 80th percentile with no convergence (even the two education scores sit 24 points apart: EA4 51st vs Privé 75th). That scatter is the result: a cognitive PRS cannot place an individual, and newer / bigger / ancestry-calibrated doesn't fix it.

On "ancestry-calibration" (the obvious next thought): the percentiles above are already ranked against an East-Asian reference (FRAPOSA) — that's the calibration. But calibration only re-ranks; the variant weights are still European, and you fundamentally cannot compare polygenic scores across ancestry groups — the once-famous European "height-selection" PRS signals were retracted as population-stratification artifacts. A genuinely EAS-derived education GWAS now exists (Chen 2023, 176k Taiwan+Korea; EAS↔EUR genetic correlation 0.87) — encouraging, but it ships as summary statistics (not a plug-in score), it's education not IQ, and even ancestry-matched it explains only 1.5–4% of years-of-schooling. EA4 (Okbay 2022, 3M; obtained via SSGAC) is now in the table above as a lead-SNP build — and it lands at the 51st percentile, no different in kind. None of these is a statement about your actual height, education, or intelligence.

Data credit (per SSGAC policy): EA4 = Okbay et al., "Polygenic prediction of educational attainment within and between families…", Nat Genet 54:437–449 (2022), PMID 35361970; summary statistics obtained from the Social Science Genetic Association Consortium (SSGAC) under its data-use agreement.

5. Ancestry & traits (interest layer)

• EAS 100% (FRAPOSA PCA vs HGDP+1kGP). Paternal haplogroup O2a1b1a2a2a1b1 (terminal SNP O-MF16703; yleaf QC-score 1.0) — a deep sub-branch of O-M122 (O2), the super-lineage carried by ~half of Han Chinese men, via the common O-002611 branch. Maternal haplogroup B4b1c1 (Haplogrep3, quality 0.90; from mtDNA — §6) — a deep East Asian subclade of B, with the classic 9-bp "Asian deletion" + 16217C confirming the B4 backbone. Both lineages East Asian; interest-layer only, no health meaning.
• Trait variants found in your data: ABCC11 T/T (dry earwax, low body odor — classic EAS), CYP1A2 A/A (fast-caffeine-metabolizer-associated), ADH1B fast-allele + ALDH2 above (alcohol metabolism), OPRM1 / DRD2-ANKK1 (opioid response / dopamine-reward).

6. Specialist callers (from the aligned 34× BAM) — gaps now resolved

The FASTQ was aligned to GRCh38, and the BAM unlocked the analyses the DRAGEN VCF couldn't:

• HLA class I (OptiType): A*02:01/*02:01, B*40:01/*40:02, C*03:04/*07:02. The three actionable alleles are all ABSENT → HLA-B*58:01 negative (no added allopurinol SJS/TEN risk despite the elevated gout PRS — allopurinol is not genetically contraindicated), B*15:02 negative (carbamazepine), B*57:01 negative (abacavir). ✅
• HLA class II (T1K on WGS): DRB1*09:01/*12:01, DQB1*03:03/*03:01, DQA1*03:02/*05:05, DPB1*05:01 (+ DRB3*01:01/DRB4*01:03). Cross-validated three ways (T1K reproduced the OptiType class I exactly; DRB1↔DRB3/4/5 linkage correct; both DR–DQ haplotypes are textbook East-Asian). Celiac disease genetically very low risk — no DQ2.5 and no DQ8 (no DQB1*02 or *03:02, cis or trans); >99% of celiac patients carry one, so its near-absence makes celiac unlikely. Not in the high-risk class II groups for type-1 diabetes or narcolepsy (no DR3-DQ2/DR4-DQ8, no DQB1*06:02). DRB1*09:01 is a modest EAS RA-risk allele (fits the higher RA PRS). Research-grade; details in output/specialist/hla2/results.md. ✅
• CYP2D6 (Cyrius): *1/*36+*10 — resolves the PharmCAT no-call. Carries the EAS-common reduced-function *10 + a non-functional *36 hybrid → activity score ~1.25 = low-normal metabolizer. Relevant to codeine/tramadol, some antidepressants/TCAs, tamoxifen (roughly standard dosing, lower end).
• SMA (SMNCopyNumberCaller): SMN1 = 2 copies → not a carrier ✅ — the single most-screened recessive condition.
• Carrier special tier (alpha-thal, Fragile X, CFTR): alpha-thalassemia (HBA) negative (no deletion; coverage full across the cluster); Fragile X (FMR1) 25 CGG = normal; CFTR poly-T 7T/7T = normal (no 5T risk allele) → not a carrier of any special-tier recessive condition ✅.
• Mitochondrial genome (10,334× on chrM): aminoglycoside-safe — reference at MT-RNR1 m.1494/m.1555 (no genetic contraindication to gentamicin etc.); no MELAS/MERRF/LHON variants; maternal haplogroup B4b1c1 (Haplogrep3 on the chrM VCF, quality 0.90 — East Asian; the 9-bp deletion + 16217C define it). ✅

• Targeted PGx (DPYD, TPMT, NUDT15, CYP2C9): reference at every key reduced-function allele (all four genes ~100% covered at 34×) → all normal function ✅ — standard fluoropyrimidine (5-FU/capecitabine) & thiopurine dosing, and the warfarin algorithm now completes (VKORC1 −1639 A/A sensitive + CYP2C9 *1/*1 normal).
• Targeted PGx round 2 (SLCO1B1, UGT1A1, NAT2) — the last No-Call/ambiguous genes, recovered from the BAM with NCBI-verified coordinates (raw genotypes in output/specialist/pgx_round2/results.md):
• SLCO1B1 — rs4149056 T/T → Normal Function ✅: no decreased-function *5 allele → standard statin dosing, no myopathy-risk genotype (reassuring given the elevated lipid PRS). Resolves the "~50 ambiguous diplotypes."
• UGT1A1 *1/*1 ✅: no *6 (rs4148323 G/G), no *28 (TATA box reads (TA)6/(TA)6, no insertion), no *80 (rs887829 C/C) → normal glucuronidation; standard irinotecan/atazanavir, no Gilbert's reduced-function allele.
• NAT2 *4/*4 → rapid acetylator ✅: none of the slow alleles (*5/*6/*7/*14 all reference) → lower risk of isoniazid hepatotoxicity, hydralazine-induced lupus, and sulfasalazine/sulfonamide hypersensitivity (those are elevated in slow acetylators).

All major callers now complete. HLA class I + II, CYP2D6, SMA, mtDNA, the full PGx panel, the carrier special tier, and the Y-haplogroup (§5) are all resolved. Remaining items are research-depth refinements only (e.g. orthogonal confirmation of the short-read SV calls; clinical-grade re-typing of anything actionable), not gaps in coverage.

Provenance & reliability

• Built from this repo's WGS pipeline: SNV/indel ClinVar/ACMG (tiered), AnnotSV SV/CNV, PharmCAT PGx, and nf-core/pgsc_calc ancestry-calibrated PRS. The old 23andMe→imputed path is fully subsumed by WGS (all 6 imputed P/LP findings re-confirmed in the WGS calls), so it was retired, not merged.
• Findings here were produced by independent extractor agents and then adversarially verified against the source files (this caught and removed, e.g., an invented SV zygosity claim).
• Limits: research-grade; short-read SV calls unconfirmed (orthogonal validation still needed); PRS magnitudes reference-sensitive; class II HLA + deep-subclade haplogroup calls are WGS-based, not clinical-grade. The PGx VCF's no-call gaps have been closed by re-genotyping from the BAM (§1/§6). Confirm anything actionable clinically.
• Privacy: this document and all underlying data are kept local and git-ignored.