Contents
Verified facts — the pinned numbers Reference

Verified facts — the pinned numbers

Every chapter in this curriculum draws its quantitative claims from this file. Nothing here was written from memory: each entry was checked against a source on the date shown.

Why this file exists. A 58-chapter textbook written over many sittings will — the gene count becomes 19,000 in one chapter and 20,000 in another, the is "47 genomes" in Chapter 43 and "200+" in Chapter 45. Pinning the numbers in one place makes that class of error impossible, and makes the whole knowledge base auditable: you can check this one file rather than re-reading 140,000 words.

How to use it when writing. Never state a number in a chapter that is not either (a) listed here, (b) derived from something listed here with the derivation shown, or (c) explicitly framed as an approximate teaching figure.

Confidence tiers

TierMeaningHow to write it
AFetched directly from the primary source (database stats page, consortium site, published paper)State the exact number, cite release/version
BConfirmed across reputable secondary sources, primary not directly fetchedState the number, hedge lightly ("approximately")
CFast-moving vendor specification or an actively contested estimateGive a range, name the date, tell the reader it moves

Human genome — annotation

Source: GENCODE Human Release 50 statistics · fetched 2026-08-13 · Tier A

QuantityValue
Protein-coding genes19,442
Long non-coding RNA genes35,885
Small non-coding RNA genes7,608
14,702
— processed10,634
— unprocessed3,535
— unitary296
— IG/TR pseudogenes237
IG/TR protein-coding segments412
Readthrough genes (protein-coding, not inside the 19,442)665
Artifact biotype19
Total annotated genes78,733
Total transcripts644,292

The four pseudogene sub-counts are components of the 14,702, not additions to it: 10,634 + 3,535 + 296 + 237 = 14,702, exactly. The IG/TR row on GENCODE's page splits into 412 protein-coding segments and 237 pseudogenes; only the 412 sit outside the four headline categories, because the 237 are already counted as pseudogenes.

The non-coding count — do not compute it by subtraction

58,195, not 59,291. The four headline categories sum to 77,637, leaving 1,096 of the 78,733 total. That residual is routinely misidentified — including in an earlier version of this file — as immunoglobulin and T-cell-receptor gene segments alone. GENCODE's own rows decompose it exactly:

412 IG/TR protein-coding segments + 665 readthrough genes + 19 artifact = 1,096

1,077 of the 1,096 are protein-coding (the IG/TR segments plus the readthrough genes), so the teaching point survives intact: subtracting protein-coding genes from the total sweeps over a thousand coding entities into the non-coding tally. Only the identification of the residual changes.

Use 35,885 + 7,608 + 14,702 = 58,195. The ~3:1 non-coding:coding ratio survives either way, but the honest figure is 58,195.

Say which 3:1 you mean. 58,195 / 19,442 = 2.99:1 only because pseudogenes are in the numerator. Transcribed non-coding RNA genes alone are 35,885 + 7,608 = 43,493, i.e. 2.24:1. A sentence whose subject is "transcribed into non-coding RNA" must therefore say better than 2:1, and reach 3:1 only by explicitly adding the pseudogenes — which are the weakest part of the anti-junk argument anyway.

Teaching note. The protein-coding count is one of the most misquoted numbers in biology. The 2001 draft-genome estimates of 30,000–40,000 have been revised steadily downward; "about 20,000" is the honest round figure and 19,442 is the current . Note also that non-coding genes now outnumber coding genes roughly 3:1 — a fact that should be used to kill the "" framing in Chapters 03 and 39.

Human genome — composition

QuantityValueTierSource
content~46% of the genome ( )BT2T repeat-element analyses, Science 2022
Sequence newly resolved by T2T-CHM13~8% of the genome, previously inaccessible — all , and the entire short arms of five acrocentric BThe complete sequence of a human genome, Science 2022
Only autonomously active human TEBas above
Alu copy number~1.1 million copies, ~11% of the genomeBAlu elements: know the SINEs
30 nm chromatin fibre compaction~40-fold cumulative relative to naked DNA — i.e. a further ~6-fold beyond the , not a further 40-foldBstandard chromatin texts
Nucleosome-occupied fraction~75% of the genome on an octamer at any moment. The remainder is mostly linker, which is not accessible; ATAC/DNase-accessible regions are ~1–3% of the genomeBENCODE data

Replication fidelity — mind the units

Tier B · DNA Replication — A Matter of Fidelity, Molecular Cell 201630140-X)

FilterError rate after this stepImprovement
Polymerase base selection~10⁻⁵—
+ 3′→5′ ~10⁻⁷10²–10³
+ ~10⁻⁹ – 10⁻¹⁰10²–10³

Final replication fidelity: ~10⁻¹⁰ per base per cell division (measured as low as 2 × 10⁻¹⁰ substitutions per base per division).

This is the single easiest unit error in the curriculum, and it appeared in four chapters and one problem set before being caught. Replication fidelity ~10⁻¹⁰ is per base per replication. The rate ~1.1–1.3 × 10⁻⁸ is per base per generation — a different quantity that sums hundreds of cell divisions and includes unrepaired chemical damage that was never a polymerase error. They differ by ~100× and are not comparable.

Sanity check, and note it overshoots: 10⁻¹⁰ × ~300 germline divisions ≈ 3 × 10⁻⁸ against a measured 1.2 × 10⁻⁸. Agreement within a factor of 2–3 is all this calculation is entitled to claim — 10⁻¹⁰ is an order-of-magnitude bound, not a constant.

Spontaneous DNA damage

ProcessRateTier
Depurination (loss of A or G by spontaneous hydrolysis)on the order of 10⁴ per cell per dayC — sources span ~5,000 to ~26,000 depending on what is counted; use the order of magnitude

Germline mutation rate

Tier B · multiple trio and long-read studies

QuantityValue
Genome-wide SNV rate~1.1–1.3 × 10⁻⁸ per bp per generation
Recent long-read pedigree estimate1.30 × 10⁻⁸ per bp per generation
Coding-sequence ratesomewhat higher, ~1.25–2.1 × 10⁻⁸
Fraction of of paternal origin~80%
Additional de novo mutations per year of paternal age~1.3–1.5

Use the rate to derive the ~60–70 de novo mutations per genome per generation figure in Chapter 16 rather than asserting it — the derivation is the teaching point.

Human generation time

Tier C · fetched 2026-08-13 · Wang RJ, Al-Saffar SI, Rogers J & Hahn MW, "Human generation times across the past 250,000 years", Science Advances 9:eabm7047, 2023 (PMID 36608127)

QuantityValue
Sex-averaged generation time, averaged over the past 250 kyr26.9 years
Paternal30.7 years
Maternal23.2 years

Curriculum convention: 27 years per generation. Use it wherever generations are converted to years, and name it in the same sentence.

Tier C because this is a model-based inference from the spectrum rather than a measurement, and it is contested; the usable literature spans roughly 25–30 years, and older sources routinely use 25 or 29. Every years-from-generations conversion inherits that ±10% spread, so quote the generation count — which is exact — and treat the years as approximate.

This was a live inconsistency, not a hypothetical. Chapter 27 converted a 69,000-generation mutation half-life at 25 years (1.7 My); Chapter 29 converted a 6,900-generation LD half-life at 28 years (194,000 y). Both now use 27 years, giving 1.9 My and 190,000 y. The choice is not cosmetic: at 25 versus 29 years that LD half-life reads 173,000 or 200,000 years — the difference between "shorter than" and "comparable to" the 315 ± 34 kyr age of the oldest H. sapiens fossils (Hublin et al., Nature 546:289, 2017).

Molecular machinery — Part 1 chapters

Fetched from the primary sources on 2026-08-13, during the adjudicated accuracy pass. Each row replaced a number that was wrong or overstated in the text.

QuantityValueTierSource
Pol III cores per active E. coli replisomethree, not the classical two — three τ molecules trimerise the polymerase. Purified holoenzyme is still often drawn as a dimerAReyes-Lamothe, Sherratt & Leake, Science 328:498, 2010
UP-element stimulation at E. coli rrnB P1~30-fold (αCTD–UP element interaction reported at 30–70-fold in vivo; a consensus UP element in the wild-type position gives 29-fold). Not tenfoldARoss et al., Science 262:1407, 1993 · Meng et al., Nucleic Acids Res. 29:4166, 2001
σ⁵⁴ activator (bacterial -binding protein) distance80–150 bp upstream of the in natural systems. Not "hundreds of base pairs"BBush & Dixon, Microbiol. Mol. Biol. Rev. 76:497, 2012
Minor (U12-type) snRNAsFour unique — U11, U12, U4atac, U6atac — plus U5, which is shared with the major spliceosome. A parts list omitting U5 is incompleteAPessa et al., PNAS 105:8655, 2008 — "U5 snRNA, in contrast, is shared between the two spliceosomes."

Lactase persistence — timing, and what it does not show

Tier A · fetched 2026-08-13 · Evershed et al., Nature 608:336–345, 2022 (Europe PMC record)

QuantityValue
Strength of selectionLP is the most strongly selected monogenic trait to have evolved over the past 10,000 years (verbatim)
Earliest European LP individual in ancient DNAc. 4,700–4,600 BC
First reaches appreciable frequencyc. 2,000 BC — "nearly three millennia after its first detection"
Milk use in Europewidespread from the Neolithic onwards, i.e. long before the frequency rise

Do not couple the sweep to the start of dairying. The paper's own model comparison rejects it: "LP selection varying with levels of prehistoric milk exploitation is no better at explaining LP trajectories than uniform selection since the Neolithic period." Population fluctuations, settlement density and wild-animal exploitation fit better, and the authors propose famine and/or increased pathogen exposure as what made lactase non-persistence costly. Write it as a very strong selection signal whose driver is still argued over — Chapter 00's worked example does exactly this.

Population reference datasets

ResourceCurrent stateTierSource
Current release is v4.1.1 (30 March 2026) — a gene-, LOFTEE- and update on v4.1 (19 April 2024), which itself fixed v4.0's -number issue. The cohort is unchanged since v4.0 (1 November 2023): 730,947 exomes + 76,215 genomes = 807,162 individuals, aligned to GRCh38, including 416,555 UK Biobank exomes. No v5 as of Aug 2026. Cite the point release — constraint metrics moved in v4.1.1 without the cohort changing.BgnomAD news index — release list fetched 2026-08-13; v4.0 announcement for cohort composition
No version number — cite the release date. NCBI ships a full weekly /XML plus a monthly archived release named by date: latest monthly ClinVarVCVRelease_2026-08 (posted 2026-08-06), latest weekly VCF clinvar_20260808 (posted 2026-08-10). "The current ClinVar" is not a citable object; a classification can change between two consecutive weeks.AClinVar FTP release listing · fetched 2026-08-13
1000 GenomesOriginal final release 2,504 unrelated individuals, 26 populations, low-. Expanded high-coverage (30×) release: 3,202 samples including 602 complete trios, NovaSeq 6000, called against GRCh38.B1000 Genomes / NYGC
HPRC Release 2 (May 2025) — 200+ individuals, 460 , ~5× expansion over Release 1. Captures >99% of common variation seen in All of Us v8. Release 1 was 47 individuals / 94 haplotypes.BHPRC Data Release 2

This corrected a stale assumption. Writing Chapter 45 from memory would have described the 2023 Release 1 draft pangenome as current. It is two years and a fivefold expansion out of date.

Sequencing platforms

Tier C throughout — vendor specifications move fast and marketing figures are best-case. Chapter 40 must give ranges, state the date, and tell the reader to check current specs.

PlatformRead lengthAccuracyNotes
Illumina NovaSeq X / X Plus2×150 bp typical (2×300 available)~Q30+, ~0.1% error25B flow cell ≈ 8 Tb per flow cell
PacBio Revio (HiFi / CCS)~15–25 kb>99.9% consensus (HiFi)~100–120 Gb per SMRT Cell in ~24 h at 15–20 kb inserts; ~70–100 Gb at 10–15 kb; ~35–70 Gb at 5–10 kb; 30 h runs at 20–25 kb
Oxford Nanopore R10.4.1 / kit V1410s of kb routine; ultra-long >100 kb achievablesimplex ~99.75% (Q26) on current high-accuracy models, against a vendor record of Q28 (99.8%); duplex ~Q30 (>99.9%)Longest reads. The accuracy gain is in the basecaller, not the chemistry — same pore and kit designation, Q20 → Q26. The duplex figure has no current vendor page behind it: treat as unpinned
Roche AXELIOS 1 (SBX)~400–600 bp in short-read (simplex) mode; up to ~1,500 bp under favourable sample and library prepnot yet independently benchmarkedLaunched 29 June 2026, research-use-only. Sequencing-by-expansion: DNA converted to "Xpandomers" read on a CMOS nanopore sensor. Same-day whole genomes claimed. Roche's own pages are not self-consistent on read length; simplex and duplex modes differ

Per-row sources, each re-fetched 2026-08-13. Tier C rows rot fastest, so they carry their own URL and date rather than a single date for the section:

RowSourceFetched
Illumina NovaSeq X / X PlusNovaSeq X Plus specifications; roadmap from the 23 Feb 2026 release2026-08-13
PacBio Reviopacb.com/revio2026-08-13
Oxford Nanoporenanoporetech.com/platform/accuracy; Q28 record from the 7 Dec 2023 release2026-08-13
Roche AXELIOS 1 / SBXSBX technology overview (400–600 bp); launch release, 29 Jun 2026 (~1,500 bp)2026-08-13

The Illumina row is deliberately conservative, not stale. Illumina's current spec page gives the 25B flow cell as ~8–10.5 Tb at 2×150; "≈ 8 Tb" sits at the bottom of the vendor's own published range, which is the right way to quote a best-case marketing figure. A February 2026 roadmap announces 25B→35B and 10B→14B flow cells over an 18-month rollout; do not write those as current specifications until a shipped datasheet says so.

This was a genuine knowledge gap. The Roche AXELIOS 1 / SBX platform launched six weeks before this curriculum was written. A chapter on sequencing technology written from memory would have presented a three-horse race (Illumina / PacBio / ONT) and missed a new entrant with a fundamentally different chemistry. Chapter 40 covers it — while being clear that independent benchmarks do not yet exist.

Single-cell droplet platform — the multiplet rate

Tier C · 10x Genomics, An introduction to GEM-X technology, 11 March 2024 · Chromium technology page · both fetched 2026-08-13

QuantityValue
Multiplet rate, current chemistry (GEM-X)~0.4% per 1,000 cells recovered
Multiplet rate, preceding chemistry (Next GEM)~0.8% per 1,000 recovered — GEM-X is quoted as a "2-fold reduction"
Maximum cells per channelup to 20,000
Median cell recovery, 3′ v4~75% (was ~60% on Next GEM v3.1)

This is the single most copy-pasted stale number in single-cell methods sections. The rate is linear in cells recovered, so it multiplies straight through into a doublet budget: at 0.4% a 10,000-cell run is ~4% doublets, at the old 0.8–1% it was 8–10%. A protocol inherited from a 2023 paper overstates the doublet burden by ~2×. Chapter 48 uses 0.4% in both §3 and the Ch 48 worked example, and states the previous figure so the reader can spot the substitution in someone else's methods.

Clinical variant interpretation

Tier B · ClinGen SVI

Chapter 55 must teach the 2015 framework as the working standard while flagging v4 as imminent. Teaching v4 as current would be wrong; teaching 2015 as unchallenged would be dated. This nuance is exactly the sort of thing a textbook written from memory gets wrong.

ClinGen Cardiomyopathy VCEP — MYH7 specification

Tier A · ClinGen Criteria Specification Registry, record GN002 · JSON fetched 2026-08-13

Record GN002, "ClinGen Cardiomyopathy Expert Panel Specifications to the ACMG/AMP Variant Interpretation Guidelines for MYH7 Version 2.0", state Released, approved 2024-04-22. Version 1.0 is Kelly et al. 2018 (PMID 29300372) and is superseded — do not quote its thresholds.

CriterionSpecification for MYH7
BA1Filtering ≥ 0.001 (0.1%) in , popmax subpopulation
BS1Filtering allele frequency ≥ 0.0001 (0.01%) in gnomAD, popmax subpopulation
PM2_Supporting≤ 0.00004 in the popmax subpopulation, using the upper bound of the 95% CI (gnomAD displays the FAF, which is the lower bound, so this must be computed)
PM1 in 167–931 (ENST00000355349 / NM_000257.4). "Rule should NOT be combined with PM5" — same-codon pathogenic variants defined the cluster, so combining double-counts
PM5"PM5 should not be combined with PM1." Where both apply at Moderate, prefer PM5 as variant-specific
PP2Not applicable for MYH7; the regional enrichment it would capture is already in PM1
PP3Approved at Supporting only — Moderate, Strong and Very Strong are all marked not applicable. REVEL ≥ 0.70. Meta-predictors preferred; individual algorithms are not independent criteria
PS2"For most cardiomyopathies, it is recommended to default to consistency: 'Phenotype consistent with gene but not highly specific'" — which on the SVI de novo point scale is 1 point (Moderate) for confirmed parentage. Shifting up or down requires stated clinical judgement

This one changed the answer to Chapter 55's flagship worked example. The chapter had applied the general ClinGen REVEL calibration (PP3_Moderate at 0.89) inside a case built on MYH7, while obeying the same panel's PM1 and PP2 rules four lines earlier. The corrected worked example applies the specification to both PP3 (down to Supporting) and PS2 (up to Moderate); the two changes cancel at 6 points, and the standing lesson is that consistency about which rulebook you are in matters more than either individual call.

Hypertrophic cardiomyopathy — gene contributions

Tier A · Sedaghat-Hamedani et al., Clin Res Cardiol 2018;107(1):30–41, PMID 28840316 · abstract fetched 2026-08-13

GeneShare of HCM cases
MYBPC320% — the largest single-gene contributor
MYH714%
TNNT2, TNNI32% each

Meta-analysis of 51 studies, 7,675 HCM patients. The ClinGen MYH7 specification uses a still lower gene contribution (10.6%, from Kelly et al. 2018).

MYH7 is not the largest contributor and its g is not 0.30. Chapter 55's worked example used g ≈ 0.30 and got AF_max = 1.2 × 10⁻⁵; at g = 0.14 it is 5.6 × 10⁻⁶, about 9 in 's ~1.6 million. Whiffin et al. 2017 build the same example on MYBPC3.

GWAS ancestry representation

QuantityValueTierSource
European-ancestry share of participants88.25% (discovery stage)AGWAS Diversity Monitor, fetched 2026-08-13, page timestamp 2026-08-07
African0.27%Aas above
African American or Afro-Caribbean2.77%Aas above
Asian6.09% · Hispanic/Latin American 1.28% · Other/mixed 1.34%Aas above
World-population share of European ancestry~16%AMartin et al. 2019, Nat Genet, PMC6563838: "~79% of all GWAS participants are of European descent despite making up only 16% of the global population"

Mind the denominator. 16% is the world share of people of European ancestry, which includes the European-descended populations of the Americas and Oceania. Europe's resident population is ~9%. Chapter 58 used 9% against an ancestry numerator, which nearly doubled the apparent over-representation ratio (5.4× → 9.7×). This is a Tier C-behaving row: the monitor moves, so re-fetch and re-date rather than copying.

Long-range familial search — Erlich et al. 2018

Tier A · Erlich, Shor, Pe'er & Carmi, Science 2018, PMC7549546 · fetched 2026-08-13

The 60% and the 2% are different results and must never be fused. Writing "~60% at 2% coverage" is a fabrication, and Chapter 58 had additionally back-solved its constant N ≈ 45 from that non-existent datum. No single N fits both published endpoints — 60% at 0.9% coverage needs N ≈ 107, >99% at 2% needs N ≈ 228 — so the independence model P = 1 − (1 − c)^N must be presented as a shape argument with an effective, conservative N, never as a fitted prediction.

Applied genomics — dated events and figures

QuantityValueTierSource
UK Generation Study200+ conditions, more than 500 genes; 97 genes / 48 conditions added April 2026; about 1 in 100 babies expected to be suspected of a condition (not diagnosed)AGenomics England — choosing conditions, fetched 2026-08-13. Supersedes the pre-April-2026 figures 462 genes / 208 conditions / 1 in 200
Genome vs exome diagnostic yieldWithin-cohort pooled 30.6% GS vs 23.2% ES, from N = 3 studies; OR 1.7, 95% CI 0.94–2.92, P = .13 — not significant. Three authors are Illumina employees/stockholdersAPandey et al., Genet Med 2025;27(6):101398, PMID 40022598, abstract fetched 2026-08-13
EU New Genomic Techniques regulationEuropean Parliament gave final approval 17 June 2026. Two categories: category 1 (equivalent to conventional breeding) follows a verification route outside the full GMO regime; category 2 stays inside it. Enters into force 20 days after OJ publication, applies two years laterBRenew Europe, 2026-06-17; Rothamsted, 2026-06-29; Appleyard Lees · fetched 2026-08-13
DOJ interim policyApproved 09.02.2019, effective 11.01.2019 (2 September and 1 November 2019), per the policy document's own page footer; publicly announced 24 September 2019ADOJ Interim Policy PDF, fetched 2026-08-13
Beacon re-identification cost250 queries against a 65-individual beacon; ~5,000 queries against a 1,000-individual beacon. Cost scales with beacon sizeAShringarpure & Bustamante, AJHG 2015;97(5):631–46, PMID 26522470, abstract fetched 2026-08-13
GEDmatch opt-in~185,000 profiles opted back in by late 2019 after the May 2019 switch to opt-out; new users choose at registration with opt-in pre-selected, and 83% stay opted inBGuerrini et al., "Four misconceptions about investigative genetic genealogy", PMC8043143, fetched 2026-08-13
Nagoya Protocol scopeDoes not cover human genetic material. Genetic resources are "any plant, animal, microbial or material of other origin"BUK BRCN guidance, fetched 2026-08-13
Utah Genetic Information Privacy ActS.B. 227, governor-signed 17 March 2021, effective 5 May 2021 — the DTC genetic-privacy wave began in 2021, not 2025–26Ale.utah.gov SB0227 (2021), fetched 2026-08-13
PCSK9Gene→hypercholesterolaemia link established 2003 via gain-of-function ADH families (Abifadel et al., Nat Genet, PMID 12730697). ARIC: nonsense in 2.6% of Black participants → 28% LDL, 88% CHD reduction; R46L in 3.2% of white participants → 15% LDL, 47% CHD reduction, HR 0.50 (95% CI 0.32–0.79), P = 0.003ACohen et al., NEJM 2006;354:1264–72, PMID 16554528, abstract fetched 2026-08-13
He JiankuiConvicted of illegal medical practice December 2019, three years' imprisonment and ¥3M fine; China's 2020 Criminal Law amendment made implantation of a gene-edited human embryo an offence carrying 3–7 yearsBcontemporaneous reporting (Xinhua, Science, STAT), checked 2026-08-13
Genetic adverse selectionHuntington carriers are up to five times as likely as the general population to hold long-term-care insurance; the authors argue modest growth in genetic information could threaten that market's viabilityBOster, Shoulson, Quaid & Dorsey, NBER w15326 (published J Public Econ 2010), fetched 2026-08-13

Still unverified — do not harden. The 85–90% concern / ~40% discrimination figures for people at risk of Huntington disease (Ch 58 §3) have no located source. RUSP condition count (Ch 57 §4) and the DPYD testing-mandate row (Ch 57 §5) were not re-checked; HRSA returns 403. The FDA/PIC PRRS pig approval (Ch 57 §7) is unconfirmed. The EU NGT row above is Tier B pending the Official Journal text — sources disagree on the vote tally, so no tally is quoted in the chapter.

Huntington disease — HTT CAG allele classes

Source: GeneReviews, Huntington Disease (NBK1305) · fetched 2026-08-25 · Tier A

QuantityValue
Normal ≤26 CAG
27–35 CAG
Reduced- HD-causing allele36–39 CAG
Full-penetrance HD-causing allele≥40 CAG
Prevalence, European ancestry9.71:100,000 (up to 17:100,000 with multi-source ascertainment)
Prevalence, East Asian and African populations0.1–2:100,000
Frequency of ≥36 CAG alleles, European ancestryas many as 1:400
Age-at-onset variance attributed to CAG lengthup to 70%
Heritable share of the residual variability10%–20%

Unit warning: all values are repeat units, not base pairs. A 40-repeat allele is 120 bp of CAG tract. Teaching note: of a disease-causing allele (~1:400) exceeds disease prevalence (~1:10,000) by about 25-fold. Nearly the whole gap is 36–39-repeat alleles that never declare themselves.

Repeat-expansion disorders — repeat, location and thresholds

Sources: GeneReviews chapters (Tier A, each fetched 2026-08-25) and the STRipy STR database (Tier B, fetched 2026-08-25)

QuantityValueTierSource
HTT repeat motif and locationCAG, coding exon 1AGeneReviews NBK1305
AR (SBMA) normal / reduced- / pathogenic≤34 / 36–37 / ≥38AGeneReviews NBK1333
ATN1 (DRPLA) normal / intermediate / pathogenic6–35 / 35–47 / 48–93AGeneReviews NBK1491
ATXN1 (SCA1) full penetrance39–44 CAG uninterrupted by CATAGeneReviews NBK1184
ATXN2 (SCA2) normal / at-risk / pathogenic≤31 / 33–34 / ≥35AGeneReviews NBK1275
ATXN3 (SCA3) normal / intermediate / pathogenic12–44 / 45–59 / ~60–87AGeneReviews NBK1196
CACNA1A (SCA6) normal / pathogenic≤18 / 20–33AGeneReviews NBK1140
ATXN7 (SCA7) normal / mutable / reduced-penetrance / full7–27 / 28–33 / 34–36 / 37–460AGeneReviews NBK1256
TBP (SCA17) normal / reduced-penetrance / full25–40 / 41–48 / ≥49BGeneReviews NBK1438 (secondary read)
DMPK (DM1) motif and locationCTG, 3′ UTRAGeneReviews NBK1165
DMPK normal / / pathogenic / congenital5–34 / 35–49 / >50 / usually >1,000AGeneReviews NBK1165
CNBP (DM2) motif, location, pathogenic rangeCCTG, 1, ~75–11,000 (mean ~5,000)AGeneReviews NBK1466
FMR1 normal / grey zone / premutation / full ~5–44 / 45–54 / 55–200 / >200 (CGG, 5′ UTR)AGeneReviews NBK1384
FXN (FRDA) motif, location, normal / pathogenicGAA, intron 1, 5–33 / 66–~1,300AGeneReviews NBK1281
C9orf72 normal / uncertain / pathogenic2–24 / 25–60 / 61 – >4,000 (GGGGCC, intron 1)AGeneReviews NBK268647
ATXN10 (SCA10) motif, location, normal / fullATTCT, intron 9, 10–32 / 800–4,500AGeneReviews NBK1175
ATXN8OS/ATXN8 (SCA8) normal / typical pathogenic15–50 / 54–250 combined (CTA·TAG)n(CTG·CAG)nAGeneReviews NBK1268
RFC1 (CANVAS) pathogenic motif and sizebiallelic (AAGGG)₄₀₀–₂₀₀₀₊, intron 2AGeneReviews NBK564656
DAB1 (SCA37) pathogenic (ATTTC)₃₁–₇₅ inserted inside an (ATTTT)nASeixas et al., AJHG 201730242-2)
FGF14 (SCA27B) pathogenic threshold≥250 GAA (250–300 reduced penetrance; ≥300 full)BPellerin et al., NEJM 2023
ZFHX3 (SCA4) motif and productGGC → polyglycine, coding; pathogenic ≥42BNat Genet 2024;56:1080–1089

Assembly note: none of these thresholds are coordinates, so no assembly applies — but any genomic coordinate for a repeat locus must name GRCh38 or and its 0-/1-based convention, because STR callers differ on both. Unit warning: STRipy and GeneReviews disagree at ATXN7, ATXN3, C9orf72, ATN1 and AR. Where they disagree, the GeneReviews value is the one recorded here.

SCA12 — PPP2R2B repeat and thresholds

Source: Srivastava et al., Brain 2017;140(1):27–36 and STRipy · fetched 2026-08-25 · Tier C — contested, give as a range

QuantityValueTierSource
Gene and locus, 5q32BHolmes et al. 1999; secondary reads
Repeat motif and positionCAG, ~133 nt upstream of the reported start site (5′ region / probable )Bsecondary reads of the PPP2R2B literature
Control repeat range4–31 (Srivastava cohort); 4–32 quoted elsewhereASrivastava et al. 2017
Classical pathogenic threshold≥51BSTRipy; long-standing convention
GeneReviews chapterNBK1202 — last updated 2011-11-17, retired 2018-12-13 ("RETIRED CHAPTER, FOR HISTORICAL REFERENCE ONLY"). While current it gave ≥51 diagnostic in the right clinical context, normal 4–32, and stated the threshold itself was "not clear", with 40–62 seen with variable, late or absent onsetBGeneReviews NBK1202 · fetched 2026-08-25
Revised lower threshold argued≥43 — "the shortest pathogenic length for the SCA12 "ASrivastava et al. 2017 (18 patients, 16 unrelated families, CAG 43–50)
Repeat length vs age at onsetPearson r = −0.65, P < 10⁻⁴, n = 124 unrelated patientsASrivastava et al. 2017
DiscoveryHolmes et al., Nat Genet 1999;23(4):391–392Bcitation verified by search; paper not fetched
Nuclear CAG RNA foci in SCA12 patient iPSC-derived neural stem cells, absent from controls; pull-down identifies 13 proteins binding the expanded repeat exclusivelydemonstrated in cells; no fraction-of-cells figure and no length threshold reportedAKumar et al. 2024, iScience 27(5):109768, PMID 38711441
at the SCA12 locusdemonstrated in cells and iPSC-derived neurons, in multiple frames — polyglutamine and polyserine from the sense strand (Delhi group), polyalanine from the antisense PPP2R2B-AS1 ORF (Hopkins group), polyleucine and polycysteine absent. Frames not reconciled between labs; no RAN product has been demonstrated in SCA12 human brainAKumar et al. 2024, iScience; Zhou et al. 2023, Mov Disord 38(12):2230–2240
Bidirectional transcription at PPP2R2B: antisense PPP2R2B-AS1 carries a CUG repeat, forms CUG RNA foci in SK-N-MC cells and induces apoptosisdemonstrated in cells; expanded transcripts reliably detected only in iPSCs and mouse models, not in post-mortem brainAZhou et al. 2023, Mov Disord 38(12):2230–2240

Teaching note: the mechanism is not settled either. The long-standing working model is cis-regulatory mis-setting of PPP2R2B expression, but RNA foci and RAN translation are now reported at this locus, and Kumar 2024 finds most PPP2R2B isoforms down in patient-derived mature neurons — which contradicts the up-regulation model the field has run on since 2010. Do not write "no toxic RNA, no translated product" for SCA12.

Teaching note: the SCA12 threshold is not settled. Tell the reader the range (43–51), name the paper that moved it, and say the field has not converged as of August 2026. A course that reports "51" as a fact is teaching a convention as though it were a measurement.

Human brain — neuron numbers

Source: Azevedo et al. 2009, J Comp Neurol 513(5):532–41 · fetched 2026-08-25 · Tier A

QuantityValue
Neurons, whole adult human brain≈ 86 billion
Neurons, ≈ 69 billion (≈ 80% of all brain neurons)
Mass, cerebellum≈ 154 g (≈ 10% of brain mass)
Neurons, cerebral cortex incl. white matter≈ 16 billion (≈ 19%)
Mass, cerebral cortex incl. white matter≈ 1,233 g
Non-neuronal cells, whole brain≈ 85 billion (glia : neuron ≈ 1 : 1)

Teaching note. The 10:1 glia-to-neuron ratio in older textbooks is wrong; this study is where the correction comes from. Note also that the cerebellum holds four-fifths of the brain's neurons in one-tenth of its mass — "cerebellar disease" is not a disease of a small structure.

Human cerebellum — cell counts (contested)

Source: Nairn et al. 1989, J Comp Neurol 290(4):527–32 · Andersen, Korbo & Pakkenberg 1992, J Comp Neurol 326(4):549–60 · Azevedo et al. 2009 · fetched 2026-08-25 · Tier C

QuantityValueTierSource
, whole 15.4 × 10⁶ (CV 19%, n = 12)CNairn 1989, fractionator
Purkinje cells, whole cerebellum30.5 × 10⁶ (CV 0.13, n = 5)CAndersen 1992, optical disector
Granule cells, whole cerebellum101 × 10⁹ (CV 0.13)CAndersen 1992
All cerebellar neurons≈ 69 × 10⁹AAzevedo 2009, isotropic fractionator
Neurons, dentate nucleus5.01 × 10⁶ (CV 0.28)AAndersen 1992
Cerebellar cortical surface area1,160 cm² (CV 0.29)AAndersen 1992

This is a genuine, unresolved disagreement. Purkinje-cell estimates differ two-fold and granule-cell estimates conflict with the whole-cerebellum neuron count. Quote as 15–30 million Purkinje cells and ~70–100 billion granule cells, and tell the reader the spread is a fact about counting methods (nucleoli vs nuclei vs homogenate), not about people.

Human Purkinje cell — morphology

Source: Masoli et al. 2024, Commun Biol 7:5 · fetched 2026-08-25 · Tier A

QuantityValue
Total dendritic length, human PC20,167 ± 15,249 µm
Total dendritic length, mouse PC2,783 ± 671 µm
Human : mouse dendritic length7.24×
Dendritic spines, human PC38,417 ± 29,210 (max 97,853)
Dendritic spines, mouse PC5,064 ± 1,342
Spine density, both species≈ 2 spines/µm
Human PCs with 2–3 primary dendritic trunks80% (mouse 10.5%)
Dendritic complexity index, human : mouse27×

Tier C caveat. A 2025 review from the same group quotes ≈ 35,000 (rodent) and ≈ 360,000 (human) spines and human dendritic lengths of 30,000–63,645 µm (Masoli et al. 2025, Front Physiol 16:1671271). Reconstruction method drives the difference. Write "tens of thousands to a few hundred thousand parallel-fibre inputs per human ".

Purkinje cell — firing

Source: Arancillo et al. 2015, J Neurophysiol 113(2):578–91 · fetched 2026-08-25 · Tier B

QuantityValue
Simple-spike rate, spontaneous in vivomean ≈ 40 Hz, range 0–200 Hz
Intrinsic pacemaker range, regular firing30–150 Hz
Complex-spike (climbing-fibre) rate≈ 0.5–1 Hz

Teaching note. are intrinsic pacemakers — they fire with all synaptic input blocked (Raman & Bean 1999, J Neurosci 19(5):1663–74). This is the load-bearing fact behind the firing-energetics account of their .

Brain energetics

Source: Attwell & Laughlin 2001, J Cereb Blood Flow Metab 21(10):1133–45 · fetched 2026-08-25 · Tier A (model) / B (body-share figures)

QuantityValueTier
Brain share of body mass≈ 2%B
Brain share of body oxygen consumption≈ 20%B
Brain share of body glucose consumption≈ 25%B
Grey-matter signalling budget: action potentials47%A
… postsynaptic effects of glutamate34%A
… resting potential13%A
… glutamate recycling3%A

Unit and scope warning. The 47/34/13/3 split is a calculated budget for rodent grey matter, not a measurement of human , and later work revised the action-potential share downward (mammalian spikes are more Na⁺-efficient than assumed) while raising non-signalling costs (Engl & Attwell 2015, J Physiol 593(16):3417–29). Use it for the qualitative claim only.

SARA — Scale for the Assessment and Rating of Ataxia

Source: Schmitz-Hübsch et al. 2006, Neurology 66(11):1717–20 · item structure from Shirley Ryan AbilityLab RehabMeasures · fetched 2026-08-25 · Tier A / B

QuantityValueTier
Items8B
Total score range0 (no ) – 40 (most severe)B
Item rangesGait 0–8 · Stance 0–6 · Sitting 0–4 · Speech 0–6 · Finger chase 0–4 · Nose–finger 0–4 · Fast alternating movements 0–4 · Heel–shin 0–4B
Validation cohorts167 and 119 SCA patientsA
Interrater reliabilityICC 0.98A
Test–retest reliabilityICC 0.90A
Internal consistencyCronbach's α 0.94A
Administration time14.2 ± 7.5 min (range 5–40)A
Correlation, Barthel Indexr = −0.80A
Correlation, UHDRS-IVr = −0.89A
Correlation, disease durationr = 0.34A
Minimal clinically important differencenot establishedC

Two warnings. (1) Gait (8) plus stance (6) supply 35% of the maximum score, so SARA is weighted toward axial function; limb-predominant change moves it little. (2) High reliability with no established MCID means a reported point-change has no independently anchored clinical meaning. Reliability is not interpretability.

Human kinases and phosphatases

Source: Manning et al. 2002, Science 298(5600):1912–34 · Shi 2009, Cell 139(3):468–84 · Olsen et al. 2006, Cell 127(3):635–48 · fetched 2026-08-25 · Tier A / B

QuantityValueTierSource
Protein kinase genes, human518 (≈ 1.7% of genes)AManning 2002
… Ser/Thr kinases428BShi 2009
… Tyr kinases90BShi 2009
Protein tyrosine genes107BAlonso 2004 / Shi 2009
Ser/Thr phosphatase catalytic subunit genes≈ 30 (PPP + PPM)BShi 2009
Phosphosite distribution Ser : Thr : Tyr≈ 86 : 12 : 2BOlsen 2006

The asymmetry is the point. 428 Ser/Thr kinases against ~30 Ser/Thr phosphatase catalytic subunits means phosphatase specificity cannot live in the catalytic subunit. It is built combinatorially by targeting subunits — which is what 's B subunits are.

PP2A — holoenzyme architecture

Source: Sandal et al. 2021, J Cell Sci 134(13):jcs248187 · Haanen, O'Connor & Narla 2022, J Biol Chem 298(12):102656 · Xu et al. 2006, Cell 127(6):1239–51 · fetched 2026-08-25 · Tier A / B / C

QuantityValueTierSource
Compositionheterotrimer: scaffold A + catalytic C + regulatory BAXu 2006
A-subunit genesPPP2R1A, PPP2R1BASandal 2021
A-subunit structure15 HEAT repeats, L-shaped superhelixAXu 2006
C-subunit genesPPP2CA, PPP2CBASandal 2021
B familiesB/B55 (PPP2R2A–D) · B′/B56 (PPP2R5A–E) · B″/PR72 (PPP2R3A–C) · B‴/striatin (STRN, STRN3, STRN4)ASandal 2021; Haanen 2022
B-subunit genes / variants15 genes, ≥ 26 transcript and splice variantsBHaanen 2022
B55 foldseven-bladed WD40 β-propeller with acidic top grooveAXu 2006
Distinct holoenzymes formable≈ 60–100C"over 60" Haanen 2022; "nearly 100" Sandal 2021
share of cellular Ser/Thr dephosphorylation≥ 50% in most cell typesBSandal 2021
PP2A as fraction of total cellular proteinup to ~1% in some tissuesBreviewed in Shi 2009

Range honesty. The holoenzyme count is a combinatorial upper bound from subunit arithmetic, not a census. Not every combination assembles, and observed abundances differ by orders of magnitude (Bβ2 holoenzymes are ~10-fold rarer than Bβ1; Dagda 2003).

PPP2R2B — gene and isoforms

Source: GTEx reference API (GRCh38/hg38) · OMIM #604326 · Dagda et al. 2003, J Biol Chem 278(27):24976–85 · Strack et al. 1998, J Comp Neurol 392(4):515–27 · fetched 2026-08-25 · Tier A

QuantityValue
ProductB55β / Bβ / PR55β, a
Location5q32; chr5:146,581,146–147,084,784 (GRCh38), minus strand (GENCODE model via GTEx; the NCBI RefSeq span differs — see — locus anatomy below)
Entrez / Ensembl5521 / ENSG00000156475
SCA12 CAG tract position133 nt upstream of the reported start site — 5′ regulatory, not coding
IsoformsBβ1 (cytosolic) and Bβ2 (outer mitochondrial membrane), differing only in the N-terminal tail
Effect of the Bβ2 N-terminus on catalysisnone — pure targeting signal; the tail alone targets GFP to mitochondria
Bβ2 : Bβ1 holoenzyme abundanceBβ2 ≈ 10-fold less abundant
Bβ2 N-terminal gating residuesSer20/21/22; phosphorylation neutralises the import signal's +3 charge and holds Bβ2 cytosolic
Tissue restriction of Bβ proteindetectable only in brain (unlike A, Bα, C)
Bβ2 regional biaspredominantly forebrain

The structural point. The SCA12 repeat is not in the ORF and makes no polyglutamine tract. SCA12 is a regulatory-dosage disease; the repeat acts as a cis element that up-regulates (Lin et al. 2010, Hum Genet 128(2):205–12).

PPP2R2B — expression (corrects a common claim)

Source: GTEx v8 median gene TPM, API ENSG00000156475.18 · Human Protein Atlas · fetched 2026-08-25 · Tier A / B

QuantityValueTier
Highest tissue, GTEx v8Brain — Frontal Cortex (BA9), 29.90 A
Brain — Cortex24.66 TPMA
Brain — 10.16 TPMA
Brain — Cerebellar Hemisphere8.76 TPMA
Median across 41 non-brain tissues0.50 TPMA
Cells — Cultured fibroblasts0.011 TPMA
HPA tissue specificitygroup enriched (brain, retina, testis)B
HPA brain-region specificitylow region specificity, detected in all regionsB

Correction. is strongly brain-enriched but not cerebellum-enriched: cerebellum ranks below every cortical and basal-ganglia region GTEx samples, and frontal cortex is ~3× higher. The cerebellar of is therefore not explained by where the gene is transcribed — it is an instance of selective neuronal vulnerability, not a solution to it.

Unknown, stated as unknown. Bulk cerebellar is dominated by granule cells, which outnumber by ~10³–10⁴. These data cannot rule out Purkinje-cell-specific enrichment; they rule out cerebellum-level enrichment. No published single-nucleus analysis of PPP2R2B by cerebellar cell type was found.

PP2A — tau phosphatase activity

Source: Liu et al. 2005, Eur J Neurosci 22(8):1942–50 · fetched 2026-08-25 · Tier A

QuantityValue
share of total tau activity, human brain≈ 71%
PP1≈ 11%
PP5≈ 10%
PP2B (calcineurin)≈ 7%
Km, tau dephosphorylation by PP1 / PP2A / PP58–12 µM (≈ intraneuronal tau concentration)
Km, PP2B5-fold higher
In Alzheimer's disease braintotal and PP2A/PP5 tau-phosphatase activity decreased; PP2B increased

Why the Km matters. With Km near the physiological substrate concentration, PP2A's flux tracks its abundance — the enzyme is neither saturated nor idle. That is the enzymological reason a dosage change in a PP2A subunit is not automatically buffered.

PP2A — DRP1 and mitochondrial fission

Source: Dickey & Strack 2011, J Neurosci 31(44):15716–26 · Merrill, Slupe & Strack 2013, FEBS J 280(2):662–73 · fetched 2026-08-25 · Tier A

QuantityValue
DRP1 inhibitory phosphositeSer637 (human) = Ser656 (rat)
Kinase (inhibits fission)PKA, anchored by AKAP1 at the outer mitochondrial membrane
(promote fission)/Bβ2 and calcineurin (PP2B)
Consequence of PP2A/Bβ2 action in neuronsmitochondria fragment and depolarise, are depleted from dendrites; dendritic outgrowth stunted, synapse number rises
Bβ2 targeting mechanismcryptic mitochondrial import signal + unfolding-resistant arrest domain, stalling the protein in the outer membrane

PP2A — regulation and inhibitors

Source: Haanen et al. 2022, J Biol Chem 298(12):102656 · Tanimukai et al. 2005, Am J Pathol 166(6):1761–71 · Junttila et al. 2007, Cell 130(1):51–62 · fetched 2026-08-25 · Tier A / B

QuantityValueTier
Methylated residue on -CLeu309 (C-terminal carboxylate)B
MethyltransferaseLCMT1; methylation permits B55 and B56 bindingB
MethylesterasePME-1; demethylation disfavours B binding and protects free C from proteasomal degradationB / A
Biogenesis chaperonesα4 (IGBP1), PTPAB
Endogenous inhibitorsI₁PP2A (ANP32A), I₂PP2A (SET) — both up-regulated in AD neocortex; I₂PP2A relocates nucleus → cytoplasm and its 39-kDa form is cleaved to ~20 kDa in ADA
Oncogenic inhibitorCIP2A — blocks PP2A toward c-Myc Ser62, stabilising c-MycA
Pharmacological inhibitorsokadaic acid, microcystin-LR (active-site binders; co-crystal structures available)A

Regulatory framing. PP2A is controlled chiefly by which holoenzyme assembles, not by dialling a catalytic rate. Methylation state, chaperone availability and subunit stoichiometry all act on assembly — which is why a change in the amount of one B subunit is a credible pathogenic mechanism.

PP2A B-subunit dosage — hypothesis, not fact

Source: reasoning from Xu et al. 2006, Sandal et al. 2021, Haanen et al. 2022, Dagda et al. 2003 · assessed 2026-08-25 · Tier C

ClaimStatus
Substrate specificity and targeting reside in the B subunitEstablished (structural + functional)
15 B-subunit genes compete for cores built from 2 A and 2 C genesEstablished (gene counts)
Holoenzyme assembly is a methylation- and chaperone-gated binding equilibriumEstablished
Therefore raising one B subunit displaces others from a limiting A–C core poolNot demonstrated quantitatively — asserted in reviews, follows from the above, but no primary measurement of the free-core pool or of endogenous B-subunit displacement at disease-relevant dosage was found
Bβ2's pro-apoptotic effect requires holoenzyme assembly (assembly-defective mutant is inert)Established (Dagda 2003) — consistent with, but not proof of, competition

Write it as a hypothesis with its logic exposed, and name the experiment that would settle it: quantitative interaction proteomics of the A subunit across a dosage series, measuring the full B-subunit occupancy distribution rather than the overexpressed subunit alone.

Somatic instability and mismatch repair in Huntington disease

Sources: GeM-HD Consortium papers and Handsaker et al. 2025 · fetched 2026-08-25

QuantityValueTierSource
GeM-HD 2015: 15 locus, effect 1hastens onset by 6.1 yearsBGeM-HD, Cell 2015;162(3):516–526
GeM-HD 2015: chromosome 15 locus, effect 2delays onset by 1.4 yearsBGeM-HD, Cell 2015;162(3):516–526
GeM-HD 2015: chromosome 8 locushastens onset by 1.6 yearsBGeM-HD, Cell 2015;162(3):516–526
GeM-HD 2019: cohort size9,064 unique HD subjects of European ancestryAGeM-HD, Cell 2019;178(4):887–900
GeM-HD 2019: modifier genesFAN1 (chr15, top locus), MSH3/DHFR (chr5), PMS1 (chr2), MLH1 (chr3), PMS2 (chr7), LIG1 (chr19), RRM2B/UBR5 (chr8) — all AGeM-HD, Cell 2019
GeM-HD 2019: the CAA resultloss of the CAACAG → earlier onset; CAACAG duplication (more glutamines) → later onset. Uninterrupted CAG length, not polyQ length, drives timingAGeM-HD, Cell 2019
Somatic threshold in striatal projection neurons≈150 CAG; SPNs below it look transcriptionally normal, above it lose identityBHandsaker et al., Cell 2025;188(3):623–639.e19
Somatic expansion and onset, human brainlonger somatic expansion in brain associates with earlier onsetASwami et al., Hum Mol Genet 2009;18(16):3039–3047
Age-dependent somatic expansion, mechanismOGG1-initiated "toxic oxidation" cycle in post-mitotic cellsAKovtun et al., Nature 2007;447:447–452

Do not quote per-locus effect sizes in years from the 2019 paper. Two automated reads of the PMC full text returned mutually inconsistent values for the same table rows, and the PMC table endpoint was CAPTCHA-gated on 2026-08-25. The 2015 figures above are safe; the 2019 per- figures must be read from Table 1 by eye before any chapter uses them.

Interruptions, premutations and reduced penetrance

Sources: GeneReviews chapters · fetched 2026-08-25 · Tier A unless marked

QuantityValueTierSource
HD reduced- penetrance (36–38 CAG)0.2%–2% across a typical life spanBGeneReviews NBK1305
SBMA: clinical prevalence vs ~1:300,000 males clinically, but 1:6,887 males carry a pathogenic expansionAGeneReviews NBK1333
FMR1 AGG interruptionsoccur about every 9–10 CGG; presence reduces risk of expansion to full for alleles <100 repeatsAGeneReviews NBK1384
FMR1 grey-zone expansion riskabout 14% of 45–54-repeat alleles expand into the range on maternal transmissionAGeneReviews NBK1384
FXTAS penetrance in male premutation carriers17% (50–59 y), 38% (60–69 y), 47% (70–79 y), 75% (≥80 y)AGeneReviews NBK1384
FXTAS penetrance in female premutation carriers >50 y16%–20% (vs ~40% in males)AGeneReviews NBK1384
ATXN1 CAT interruptionsa 36–44 CAG allele with CAT is normal; without CAT it is mutable or pathogenicAGeneReviews NBK1184
ATXN10 ATCCT interruptionassociated with higher prevalence of epileptic seizures; interrupted alleles show with paradoxical repeat contractionAGeneReviews NBK1175
SCA8 penetrancereduced penetrance at repeats of all sizes; alleles 71 – >1,300 occur in affected and unaffected peopleAGeneReviews NBK1268
DM2 –no significant correlation between CCTG size and age of onset or severity; anticipation not confirmedAGeneReviews NBK1466
SCA6 anticipationnot observedAGeneReviews NBK1140
DRPLA anticipationoffspring onset ~26–29 y earlier than affected fathers, ~14–15 y earlier than affected mothersAGeneReviews NBK1491
DM1 transmitted expansion sizepaternal median 425 (range 70–2,000); maternal median 200 (range 57–1,400)AGeneReviews NBK1165
FRDA onset by GAA1 lengthGAA1 <700 → mean onset 18 y; >700 → mean onset 9.7 y; <500 → onset >25 y; <300 → onset >40 yAGeneReviews NBK1281
FRDA carrier frequency1/60–1/100AGeneReviews NBK1281
C9orf72 share of ALS/FTD30–50% familial ALS, 4–10% sporadic ALS, ~25% familial FTDAGeneReviews NBK268647
RFC1 share of adult-onset 14%–22%; 82%–97% of full-CANVAS phenotypesAGeneReviews NBK564656
SCA3 share of ataxia families20%–50% worldwide; 21%–25% in the US and CanadaAGeneReviews NBK1196

Teaching note: "premutation", "" and "reduced-penetrance allele" are three different things. An intermediate allele is unstable but harmless to its carrier; a premutation (the FMR1 55–200 case) causes a different disease by a mechanism the full mutation cannot use; a reduced-penetrance allele (HD 36–39) is the disease allele in someone who may never manifest.

SCA12 — locus anatomy

Source: STRchive SCA12_PPP2R2B, STRchive-loci.json · fetched 2026-08-25 · Tier A

QuantityValue
Gene
Cytoband5q32
Gene strandminus
Repeat, GRCh38chr5:146,878,728–146,878,759
Repeat, GRCh37chr5:146,258,291–146,258,322
Repeat, chr5:147,414,734–147,414,765
Reference copy number10.7
Motif, plus/reference strandGCT (i.e. the tract reads CTG-like on the plus strand)
Motif, gene/sense strandAGC (i.e. the tract reads CAG-like on the mRNA)
Region 5′ UTR
Gene span, GRCh38chr5:146,580,742–147,081,520
RefSeq transcript variants10

Unit and strand warning. (CAG)n, (CTG)n, (AGC)n and (GCT)n in the literature all name the same ~32-bp reference element (30 bp as the ExpansionHunter catalog bounds it, 32 bp as STRchive does). PPP2R2B is on the minus strand, so the disease-relevant CAG reading is on the mRNA and the plus-strand genome browser shows CTG; a trinucleotide repeat has no canonical , so databases differ by a cyclic rotation. Never infer a different locus from a different motif spelling.

SCA12 — allele ranges (sources disagree)

QuantityValueTierSource
Normal 4–32BGeneReviews chapter (retired 2018-12-13), fetched 2026-08-25
Normal alleles6–32ASTRchive JSON, fetched 2026-08-25
Normal alleles4–31ASrivastava et al. 2017, Brain 140(1):27–36
Benign call anchor7–28A RCV000005966, NM_181675.3(PPP2R2B):c.27CAG[(7_28)], fetched 2026-08-25
40–49ASTRchive JSON, fetched 2026-08-25
"Intermediate pathogenic"43–50ASrivastava et al. 2017, Brain 140(1):27–36
Pathogenic alleles51–78ASTRchive JSON; GeneReviews (retired) gives ≥51 as diagnostic
Proposed lowered threshold≥43ASrivastava et al. 2017, Brain 140(1):27–36
Shortest reported pathogenic allele46ADong, Wu & Wu 2015, Parkinsonism Relat Disord 21(4):398–401
Shortest reported symptomatic alleles40 and 42AGanaraja et al. 2022, Tremor Other Hyperkinet Mov 12:13
Index-family expanded alleles66–78 (fully penetrant)BGeneReviews SCA12 chapter (retired)
Indian founder expanded alleles51–69ABahl et al. 2005, Ann Hum Genet 69(Pt 5):528–534
Non-penetrant / very-late-onset carriersreported at 45–62BGeneReviews SCA12 chapter (retired)
Unaffected carriers within an affected familyCAG-39ASrivastava et al. 2017, Brain 140(1):27–36

Teaching note. ≥51 CAG with a compatible is diagnostic and ≤31 is normal — everything from 32 to 50 is contested territory, and the boundary has moved downward twice since 2015 (46, then 43, then symptomatic cases at 40). Report a 44-repeat allele as "in a range where pathogenicity is reported but not established", never as "negative".

SCA12 — epidemiology

QuantityValueTierSource
Fraction of Indian referrals (pan-India, ~5,600 patients, 10 yr)8.6% (490)ASharma et al. 2022, Adv Genet 3(2):2100078
SCA2 in the same cohort, for comparison8.5% (482)ASharma et al. 2022
Fraction of AD ataxia at AIIMS North India~16% (20/124)ABahl et al. 2005, Ann Hum Genet 69(Pt 5):528–534
Agarwal ancestry in a 49-patient Indian cohort79.6% (10.2% explicitly non-Agarwal)AGanaraja et al. 2022, Tremor 12:13
association with expanded P = 0.000, 20 families, markers spanning ~137 kbABahl et al. 2005
Independent origins of the expansion≥2 (Indian founder haplotype absent from the American )ABahl et al. 2005
Yield in a European ataxia 1 family / 247 index casesAFujigasaki et al. 2001, Ann Neurol 49(1):117–121
expansions in familial essential tremor probands1 / 515 (of 6/515 pathogenic STR expansions total)AZhou X et al. 2024, Brain Commun, PMID 38961870
Population prevalence per 100,000no published estimate—STRchive prevalence: null; absent from GeneReviews and Orphanet

Teaching note. "Second commonest SCA in India" is the familiar line and comes from a 124-family AIIMS series. The largest series (Sharma 2022, ~5,600 referrals) puts fractionally ahead of SCA2 — 8.6% vs 8.5%, a margin inside the noise. Say "SCA12 and SCA2 are jointly the commonest SCAs in Indian referral cohorts", and never convert a referral fraction into a population prevalence.

SCA12 — clinical phenotype

Source: Ganaraja et al. 2022, Tremor Other Hyperkinet Mov 12:13 (n = 49 from 42 families) · fetched 2026-08-25 · Tier A

QuantityValue
Tremor at presentation95.9%
at presentation73.5%
Head tremor55.1%
Voice tremor42.8%
Impaired tandem gait89.8%
Dysmetria75.5%
Dysarthria57.1%
Cognitive dysfunction22.4%
Psychiatric disturbance8.1%
Mean age at onset46.38 ± 11.7 yr
Mean expanded repeat length53.26 ± 6.10
Positive family history93.8%
MRI: cerebellar atrophy only34.7%
MRI: cerebral atrophy only16.3%
MRI: both34.7%
MRI: normal6.1%
QuantityValueTierSource
Cognitive impairment (dedicated non-motor study, n=34)61.76%ABasu et al. 2024, Front Neurol 15:1464149
Psychiatric disorders (n=21)67%AChoudhury et al. 2018, Mov Disord Clin Pract 5(1):39–46
Tremor as first symptom (n=21)90%AChoudhury et al. 2018
Mean age at onset (n=21)51.33 ± 8.98 yr, range 41–69AChoudhury et al. 2018
Repeat length vs age at onsetr = −0.760, P = 0.0001AChoudhury et al. 2018
Repeat length vs age at onsetno significant correlationAGanaraja et al. 2022
Typical onset window26–50 yr; full range 8–62 yrASTRchive JSON
Index-family onset"fourth decade", mean 34–38 yrBGeneReviews chapter (retired)
Life expectancybelieved not shortened; no survival study publishedCNational Ataxia Foundation, fetched 2026-08-25

Teaching note. The two Indian cohorts disagree about whether repeat length predicts onset age (r = −0.76 vs none) and the index family looks a decade or two younger at onset than any Indian series. Both disagreements are live. Quote a cohort, never a consensus.

Diagnostic warning. 6.1% of a genetically confirmed SCA12 cohort had a normal MRI, and two short-duration patients in a second cohort had no atrophy. Imaging supports the diagnosis; it never excludes it.

SCA12 — anticipation and instability (contested)

QuantityValueTierSource
"Insufficient evidence for anticipation"AGeneReviews Hereditary Overview, last revised 2026-07-09, fetched 2026-08-25
Anticipation"a moderate degree of anticipation has been observed"BGeneReviews chapter, retired 2018-12-13
instability"modestly unstable, length variations of a few triplets among sibship members"BGeneReviews SCA12 chapter (retired)
Parent of origin (n=21)7 maternal, 7 paternal, 1 biparental — no bias detectedAChoudhury et al. 2018
Somatic in braindetected across regions; showed the least instability, with higher methylation and lower expressionAParthaje et al. 2025, Cerebellum 24(3):60 — n = 1 brain
Biallelic carriersCAG-45/45 and CAG-42/51; no difference from heterozygous CAG-51ASrivastava et al. 2017, Brain 140(1):27–36

Teaching note. SCA12 is the counter-example to "repeat disease ⇒ anticipation ⇒ paternal bias." Current GeneReviews says the evidence for anticipation is insufficient, the only parent-of-origin dataset is 7-vs-7 in 21 patients, and the single brain studied showed least in the cerebellum — the opposite of the striatal pattern in Huntington disease. The mechanistic rules learned from HTT and DMPK do not transfer automatically.

SCA12 — mechanism evidence (nothing established)

ClaimStatusTierSource
Repeat region has activity that increases with repeat lengthdemonstrated in reporter assaysAO'Hearn et al. 2015, Mov Disord 30(13):1813–1824
Repeat acts as a cis element up-regulating ; CREB1/SP1 up, TFAP4 downdemonstrated in cellsALin et al. 2010, Hum Genet 128(2):205–212
Expansion raises the 7B7D transcript and Bβ1 protein, and yields a polyserine tract that triggers apoptosisdemonstrated in cell modelsAZhou et al. 2024, Mov Disord 39(10):1886–1891
Most PPP2R2B isoforms are down-regulated in patient-derived mature neuronsdemonstrated — contradicts the overexpression modelAKumar et al. 2024, iScience 27(5):109768
Bβ2 targets to mitochondria and promotes apoptosisdemonstrated (overexpression)ADagda et al. 2003, JBC 278(27):24976–24985
Bβ2 dephosphorylates Drp1 S637, drives fission, antagonises neuronal survival; silencing Bβ2 is protectivedemonstrated (overexpression/knockdown)ADagda et al. 2008, JBC 283(52):36241–36248
Drosophila ppp2r2b overexpression → neurodegeneration, mitochondrial fragmentation, ROS; rescued by antioxidants and SOD2demonstrated in flyAWang et al. 2011, JBC 286(24):21742–21754
The expansion specifically raises Bβ2 in patient neuronsnot demonstrated — the repeat is 5′ to the Bβ1 first exon, not Bβ2's—inference from Dagda 2003 + Zhou 2023
Nuclear CAG RNA foci in patient iPSC-derived neural stem cells; 13 proteins bind the expanded repeat exclusivelydemonstratedAKumar et al. 2024, iScience
CUG RNA foci from the antisense transcript PPP2R2B-AS1demonstrated in SK-N-MC cellsAZhou et al. 2023, Mov Disord 38(12):2230–2240
at the SCA12 locus — polyQ and polySer (sense, Delhi group); polyAla (antisense, Hopkins group); polyLeu and polyCys absentdemonstrated in cells and iPSC neurons; frames not reconciled between labsAKumar 2024; Zhou 2023
Any RAN product in SCA12 human brain tissuenever shown—negative: Zhou 2023 could not reliably detect expanded transcript in one post-mortem brain
Tau dephosphorylation as an SCA12 mechanismproposed in a 2001 review, never tested in any SCA12 system; the one autopsy series found tau-negative inclusions—Holmes et al. 2001, Brain Res Bull 56(3–4):397–403; O'Hearn et al. 2015
Autopsy findingsenlarged ventricles, marked cerebral cortical atrophy, loss, less prominent cerebellar/pontine atrophy, ubiquitin-positive intranuclear inclusions negative for polyQ, α-synuclein, tau and TDP-43AO'Hearn et al. 2015 — a handful of brains
Overall mechanism confidenceSTRchive gene–disease evidence score 8.5 / 18 ("Moderate"); mechanism "incompletely established"ASTRchive curation page, fetched 2026-08-25

Teaching note. SCA12 is the honest counterweight to the tidy repeat-disease narrative. Twenty-seven years after the gene was found, the field has five plausible mechanisms — promoter-driven overexpression, Bβ2-driven mitochondrial fission, sense RNA toxicity, antisense RNA toxicity, and RAN translation in at least three frames — supported almost entirely by overexpression constructs, cell lines, one fly model, iPSC neurons and a handful of autopsy brains. Two of the strongest results point in opposite directions (Bβ1 up in Zhou 2024, isoforms down in Kumar 2024). "We do not know" is the correct answer, and it is not a gap in the teaching — it is the lesson.

SCA12 — therapy and trial readiness

QuantityValueTierSource
Published ASO or RNA-targeting therapeutic for noneAdirected Europe PMC search, 2026-08-25, zero hits
Disease-modifying clinical trial in none foundAsame search
Randomised trial of propranololn = 60, extended-release propranolol to 240 mg/day vs placebo, 8 weeks; significant reduction in TETRAS PS, improved ADL, SARA and SF-36; no safety discontinuationsAMohapatra et al. 2026, Mov Disord 41(2):373–383
Longitudinal progression data3 patients over 5–6 yr; gait variability (step length, stance time, step time) progressedASiddique et al. 2021, Clin Park Relat Disord 5:100102
Objective motor biomarkerML on reaching kinematics separates SCA12 from essential tremor at 83.3% accuracyABayen et al. 2026, 25(4):92
Candidate plasma biomarkerAβ40 decreased, Aβ42/Aβ40 ratio increasedABanerjee et al. 2026, Cerebellum 25(3):75
Candidate blood transcript biomarker5 mitochondrial quality-control genes down in PBMCsAAnsari et al. 2026, Parkinsonism Relat Disord 145:108228
Multi-site prospective natural-history cohort with annualised SARA ratesnone foundAdirected search, 2026-08-25

Teaching note. SCA12 in 2026 has a positive randomised trial — of a beta-blocker for tremor. That is symptom control, not disease modification. The gap between "we can measure this disease well enough to run an RCT" and "we have anything to give it" is the entire distance between a trial-ready and a druggable mechanism.

SCA12 locus — genomic definition (GRCh38)

Source: Illumina ExpansionHunter hg38 variant catalog · fetched 2026-08-25 · Tier A

QuantityValue
Catalog entry (0-based half-open)chr5:146878727-146878757, LocusStructure (GCT)*
Same interval, 1-based (samtools form)chr5:146,878,728–146,878,757 (GRCh38)
Reference 10 repeat units (30 bp)
v4 API main_reference_region (cross-check)chr5:146878727-146878757 · fetched 2026-08-25
STRchive interval (cross-check)chr5:146878727-146878759 · strchive.org/loci/sca12_ppp2r2b

The catalog motif GCT is the plus-strand reading; is transcribed from the minus strand, where the repeat reads CAG. STRchive annotates the repeat as 5'UTR; the gnomAD v4 API annotates the same interval "coding: polyserine" against a different transcript — which transcript the repeat is "in" is -dependent. Teach the disagreement.

Repeat-genotyping tools — versions and platforms

Source: GitHub releases API + bioconda package API, per-tool repos (ExpansionHunter, REViewer, ExpansionHunterDenovo, STRling, GangSTR, TRGT) · fetched 2026-08-25 · Tier A

QuantityValue
ExpansionHunter latest releasev5.0.0 (2021-08-20); bioconda incl. osx-arm64
Loci in default hg38 catalog31, including
REViewer latest releasev0.2.7 (2022-02-02); repo archived; bioconda osx-64/linux only
ExpansionHunter Denovo latestv0.9.0 (2020-03-04); bioconda linux only
STRling latestv0.6.0 (2025-12-06); bioconda linux-64 only
GangSTR latestv2.5 (2021-01-29); bioconda incl. osx-arm64
TRGT latestv5.1.0 (2026-06-10); bioconda linux-64 only

Versions move; re-check dates before republishing. The archived status of REViewer is a teaching point about clinical-bioinformatics tool lifecycles, not a reason to drop it.

Lab-11 public data — URLs, sizes, executed results

Source: HEAD requests + executed samtools/ExpansionHunter runs on this machine · 2026-08-25 · Tier A

QuantityValue
HG00096 30x (AWS mirror)15.7 GB — https://1000genomes.s3.amazonaws.com/1000G_2504_high_coverage/data/ERR3240114/HG00096.final.cram
Its .crai index1.4 MB
Reads in chr5:146877000-146880000 (executed)888, all 150 bp; region 145 KB
ExpansionHunter of HG00096 at (executed)10/14 GCT units (CI 10-10/14-14; 12+14 spanning reads; 0 IRRs)
chr5 reference slice via EBI CRAM registry (REF_PATH)173 MB cached, fetched automatically by MD5
Full GRCh38_full_analysis_set_plus_decoy_hla.fa (alternative)3.26 GB
HG002 HiFi GRCh38 BAM119.8 GB (bai 23.8 MB) — region extraction: 69 reads, mean 14,552 bp, 9 s
HG002 source materialDNA extracted from the EBV-immortalised B-lymphoblastoid cell line GM24385 (Coriell) — NIST RM 8391 certificate · Tier A
HG002 repeat by direct count (executed)53 spanning HiFi reads, modal 10 CTG units (≈10/10)
GIAB HG002 ONT-UL GRCh38 BAM187.5 GB (bai 54.6 MB) — 41 reads over the repeat, counted in 11 s
Whole-lab download budget≈300 MB

samtools view -c <URL> region works with no reference (counting never decodes bases); extracting sequences from CRAM without REF_PATH fails with Unable to fetch reference — run the failure first, then set REF_PATH=...:https://www.ebi.ac.uk/ena/cram/md5/%s. Cohort: 3,202 samples at 30x, NovaSeq 2×150 bp (Byrska-Bishop 2022, Cell 185:3426).

PPP2R2B expression — GTEx v10

Source: GTEx v10 gene median TPM file (downloaded, 8.8 MB) and GTEx portal API v2 · fetched 2026-08-25 · Tier A

QuantityValue
Genes × tissue columns in the median- file59,033 × 68 (54 primary tissues + 14 LCM sub-columns)
median TPM, Brain_Frontal_Cortex_BA929.71 (highest tissue)
PPP2R2B median TPM, Brain_Cerebellum / Cerebellar_Hemisphere10.02 / 8.78
PPP2R2B median TPM, Whole_Blood / Liver / Muscle_Skeletal0.40 / 0.05 / 0.04
Annotated transcripts returned by the transcript API21; top cerebellar isoform ENST00000504198.5 at 6.26 TPM
Full transcript-TPM matrix (do NOT assign)4.34 GB; no median-transcript flat file exists in the v10 bucket

GTEx v10 released Nov 2024 (AnVIL announcement); ~12% more samples than v8 (Tier B). Numbers above are v10 and will change with any future release — name the version wherever they appear.

Lab-12 single-cell data — Siletti human brain atlas

Source: CZ CELLxGENE curation API, collection 283d65eb-dd53-496d-adb7-7570c7caa443 · fetched 2026-08-25 · Tier A

QuantityValue
AtlasHuman Brain Cell Atlas v1.0 (Siletti et al. 2023, Science 382:eadd7046)
Cerebellar Vermis (CBV) dissection h5ad71,874 cells, 431 MB, direct download, no registration
lateral hemisphere (CBL) h5ad28,028 cells, 160 MB
Supercluster: Cerebellar inhibitory h5ad14,411 cells, 150 MB

Direct URLs are content-addressed (datasets.cellxgene.cziscience.com/<id>.h5ad) and may rotate on dataset revision — re-resolve via the collection page if a 404 appears. The zero-install fallback is the cellxgene Explorer on the same collection.


What this review changed

Recorded so the corrections aren't silently re-introduced later:

  1. — Release 1 (47 genomes) is superseded by Release 2 (200+ individuals, 460 , May 2025).
  2. Sequencing platforms — Roche AXELIOS 1 / SBX launched June 2026 and belongs in the platform comparison.
  3. ACMG — v4 exists in draft; the 2015 guidelines need to be framed as "current but under active revision", not as settled.
  4. Gene count — pinned to the exact current (19,442, GENCODE 50) rather than a remembered round number, with the non-coding:coding ratio available to make a pedagogical point.
  5. UP element (2026-08-13) — ~30-fold, not tenfold, at rrnB P1. The old figure understated it 3×.
  6. σ⁵⁴ activators (2026-08-13) — 80–150 bp upstream, not "hundreds of base pairs".
  7. Pol III stoichiometry (2026-08-13) — three cores per live replisome; "dimeric holoenzyme" is the pre-2010 picture.
  8. Minor (2026-08-13) — U5 belongs in the parts list; it is shared with the major spliceosome.
  9. Lactase persistence (2026-08-13) — the sweep is not coupled to the onset of dairying; the stays rare for nearly three millennia after first appearing.
  10. GENCODE residual (2026-08-13) — the 1,096 genes outside the four headline categories are 412 IG/TR coding segments + 665 readthrough + 19 artifact, not "IG/TR gene segments" alone. 1,077 are protein-coding, so the do-not-subtract lesson stands; the identification was wrong. The 237 IG/TR are inside the 14,702, not additional to it.
  11. ONT accuracy (2026-08-13) — simplex ~Q26 (99.75%), vendor record Q28. The old ~Q23 ceiling was ~2.5 years stale on its own pinning date. The duplex ~Q30 figure now has no vendor page behind it.
  12. PacBio Revio throughput (2026-08-13) — ~100–120 Gb per SMRT Cell in ~24 h, not ~60–90 Gb in ~30 h. One chemistry generation behind.
  13. Roche SBX read length (2026-08-13) — ~400–600 bp short-read mode, up to ~1,500 bp; the pinned "~175 bp" was a pre-launch example figure and appears nowhere in Roche's shipped documentation.
  14. 10x multiplet rate (2026-08-13) — ~0.4% per 1,000 cells recovered on current chemistry, roughly half the 0.5–1% previously pinned. Worked examples in Chapter 48 recomputed.
  15. Human generation time (2026-08-13) — newly pinned at 27 years. It was unpinned, and the curriculum was silently using 25 (Ch 27), 28 (Ch 29) and 29 (problem set 9) for the same conversion. Chapters 27 and 29 now agree; the question banks and problem sets still need reconciling.
  16. ClinGen MYH7 VCEP (2026-08-13) — v2.0 (approved 2024) supersedes the Kelly 2018 v1 thresholds. PP3 is approved at Supporting only for MYH7, so Chapter 55's flagship worked example was applying the general REVEL calibration inside a case that obeys the specification elsewhere. PS2's specification default runs the other way (Moderate, not Supporting). Both now applied; the tally is unchanged at 6 points and Likely pathogenic, and the standing lesson is check for a specification before deriving your own thresholds.
  17. HCM gene contribution (2026-08-13) — MYBPC3 (~20%) is the largest single-gene contributor, not MYH7 (~14%). Chapter 55's g ≈ 0.30 was unsourced and wrong under the chapter's own definition of g; AF_max for HCM via MYH7 is 5.6 × 10⁻⁶ (~9 alleles in 1.6M), not 1.2 × 10⁻⁵ (~19).
  18. Erlich et al. 2018 (2026-08-13) — "~60% of US individuals of European descent at ~2% " fuses two different results and states a figure the paper does not contain. 60% is the hit rate for a 1.28M-record database (~0.9% coverage); at 2% coverage the paper projects >99%. Chapter 58 had back-solved its N ≈ 45 from the fused datum; the model is now presented as a shape argument with an explicitly effective, conservative N.
  19. ancestry denominator (2026-08-13) — European ancestry is ~16% of the world (Martin et al. 2019); 9% is Europe's resident population, and using it against an ancestry numerator inflated the over-representation ratio from ~5.4× to ~9.7×. Monitor figures re-fetched live: 88.25% European, 0.27% African.
  20. Nagoya Protocol (2026-08-13) — does not cover human genetic material. Listing it among remedies for the Havasupai structure of harm was a category error; Chapter 58 now says so explicitly rather than deleting it, because the mis-citation is common.
  21. Generation Study (2026-08-13) — >500 genes / 200+ conditions / ~1 in 100, superseding 462 / 208 / 1 in 200. Corrected in Ch 57 §4 and Ch 54 (Ch 54's "more than 200 conditions" was already right and was left alone).
  22. EU NGT regulation (2026-08-13) — adopted 17 June 2026. Chapter 57's "the EU still regulates most such organisms under its 2001 GMO directive" became false, and the section's "one technique, three regulatory philosophies" punchline went with it — the EU's two-tier, edit-based structure now parallels the UK Act.
  23. DOJ forensic-genealogy policy (2026-08-13) — Ch 58 said September 2019, Ch 57 said November 2019, and both were half-right: approved 2 September, effective 1 November. Both chapters now name the event rather than a bare date.

Both question-bank propagations of items 17 and 18 were resolved on 2026-08-13. question-banks/qb-part-12.md no longer back-solves N from the fabricated "60% at 2% coverage" datum; it now states the model as a shape argument and gives Erlich's two irreconcilable fits (N ≈ 107 at ~0.9% coverage, N ≈ 228 at 2%), matching Ch 58. question-banks/qb-part-11.md now carries 5.6 × 10⁻⁶ / ~9 alleles and names MYBPC3 as the largest single-gene contributor, matching Ch 55.

Re-verification

Everything above carries a verification date of 2026-08-10, except rows and sections explicitly dated 2026-08-13 (added during the adjudicated accuracy pass). The Tier C rows (sequencing platforms) and the ACMG row are the ones that will rot first. Re-check them before relying on this curriculum a year from now.

The 2026-08-13 pass found that prediction was already too optimistic. Two of the four sequencing-platform rows and the single-cell multiplet rate were stale on their original 2026-08-10 pinning date — the ONT figure by about two and a half years. A vendor row is not verified because someone wrote a date next to it; it is verified because someone opened the vendor's page. That is why the Tier C rows now carry their own per-row URL and fetch date: the next person to re-check them can see exactly what was opened and when, and does not have to trust a section-level date stamp.