Contents
Structural verification report Reference

Structural verification report

Run date: 2026-08-13 · repo root: the repository root

Every figure below came from a command executed against the working tree on the date above. .venv/, .git/ and .gstack/ are excluded from every scan. 97 markdown files were in scope.

Headline. One broken internal link (README.md → GLOSSARY.md, a file that has not been written yet). Three labs are missing against the advertised count of eleven. Everything else — chapter completeness, section structure, numbering, artifact hygiene, pinned numbers, problem sets, question banks — passes clean.


1. Chapter completeness

PASS. All 59 chapters (00–58) are present across part-00 … part-12, and every one of them carries all six required sections.

chapters found: 59
missing numbers: []
duplicate numbers: []
min/max: 0 58

--- section gaps ---
chapters with gaps: 0

The six sections checked, as headings at any level, case-insensitively, with curly apostrophes normalised to straight:

Distribution across parts:

PartChaptersFiles
part-00-orientation00–012
part-01-molecular-foundations02–087
part-02-transmission-genetics09–157
part-03-genome-instability16–205
part-04-gene-regulation21–255
part-05-population-genetics26–294
part-06-quantitative-genetics30–323
part-07-molecular-evolution33–353
part-08-methods36–383
part-09-genomics39–457
part-10-functional-genomics46–505
part-11-human-and-statistical-genomics51–566
part-12-applications-and-ethics57–582
Total00–5859

A Python checker walked every .md file, skipped fenced code blocks, extracted both inline links (bracketed text immediately followed by a parenthesised target) and reference definitions of the form [id]: target, discarded http(s):, mailto:, ftp:, data: and bare-anchor targets, stripped #anchors, URL-decoded the remainder, resolved it against the containing file's directory with os.path.normpath (so ../ is handled properly) and tested os.path.exists.

markdown files scanned: 97
total links found: 1870
relative links checked: 1831
BROKEN: 1

README.md:180	target=GLOSSARY.md	resolves_to=GLOSSARY.md

1 broken link out of 1,831 relative links checked. Every one of the several hundred cross-part chapter links — the ../part-NN-.../NN-slug.md form used throughout the Connections sections, which is exactly where between independently written chapters would show up — resolves correctly.

The single failure is README.md:180:

- [Glossary](GLOSSARY.md) — every term, defined once

GLOSSARY.md does not exist at the repo root. This is not a typo with a nearby correct target, so it was not auto-fixed — see Needs attention.


3. Numbering contiguity

PASS. Chapter numbers run 00–58 with no gaps and no (see the script output in §1). 59 numbered chapter files, 59 distinct numbers, min = 0, max = 58.


4. Artifact sweep

PASS on all three checks.

Stray generation markers

$ grep -rn --include='*.md' -E '<invoke|<parameter name|antml:' . \
    --exclude-dir=.venv --exclude-dir=.git --exclude-dir=.gstack
RESUME.md:119:**Generation artifacts.** Fix agents found stray `<invoke name="...">` text and malformed
RESUME.md:124:grep -rn "<invoke\|<parameter name\|antml:" --include="*.md" .

Both hits are in RESUME.md, which documents the pattern deliberately, and both are inside inline-code spans. Zero real artifacts in the 96 other files.

Note for future sweeps: this report is now a third deliberate-mention file, alongside RESUME.md. Exclude reference/verification-report.md as well, or the grep will report itself.

<details> / </details> balance

A naive count reports three mismatches — README.md (1/0), STUDY-GUIDE.md (1/0) and RESUME.md (2/1). All are false positives: every occurrence is an inline-code mention of the tag in prose, e.g.

README.md:62:      | 11 sets, worked solutions folded in `<details>` |
STUDY-GUIDE.md:54: Solutions are folded into `<details>` blocks. This is deliberate.

Re-running with inline-code spans stripped reports files containing real <details>: 79 and mismatched (excluding inline-code mentions): 0.

79 files use real <details> blocks and all 79 balance.

Trailing newline

files without trailing newline: 0

All 97 files end in a newline. Nothing to fix.


5. Pinned-number consistency

Checked against verified-facts.md. PASS on all five.

Pinned figureVerdictEvidence
19,442 protein-coding genesConsistent27 occurrences across 12 files (ch 00, 03, 06, 08, 22, 24, 37, 39, 44, 54, qb-01, qb-09, ps-06, formulas.md, verified-facts.md). No competing exact count anywhere.
58,195 non-coding genesConsistent12 occurrences; every one either states it directly or shows the derivation 35,885 + 7,608 + 14,702 = 58,195.
59,291 must appear nowhere as factSatisfied4 occurrences, all four explicitly flagging it as the wrong value — see below.
~46% Consistentch 00 L459, ch 03 L237/L459, ch 19 L18, ch 35 L360, ch 39 L223/L352. No competing TE figure.
730,947 exomes / 76,215 genomesConsistentch 55 L224 states both plus the 807,162 total; ch 51 L371, ch 58 L383 and ch 54 L137 cite 807,162. No other gnomAD size claim.
HPRC Release 2, 460 Consistentch 45 L175/L217/L295, qb-09 L263, ch 58 L383. Nothing describes Release 1 as current.

The 59,291 hits, in full

All four are the number being named in order to be rejected:

part-09-genomics/39-genome-landscapes.md:202: entities into the non-coding tally and returns 59,291 instead of the correct 58,195.
part-09-genomics/44-annotation.md:302:        obtained as 78,733 − 19,442 = 59,291, which quietly reclassifies the 1,096 IG/TR segments as
question-banks/qb-part-09.md:231:            A: Because 35,885 lncRNA + 7,608 small ncRNA + 14,702 pseudogenes = 58,195, whereas subtracting …
reference/verified-facts.md:41:                 **58,195**, not 59,291. The four categories above sum to 77,637, …

This is the correct treatment: the wrong number is taught as wrong.

HPRC Release 1 mentions, in full

Three chapters mention Release 1. None presents it as current:

Two non-contradictions worth recording

Both were checked and neither is an error:

  1. part-01/03-genomes-chromosomes-chromatin.md:268 — a mermaid node reads repetitive DNA ~50%+ of the human genome. This is repetitive DNA, a broader category than TE-derived sequence, and L290 of the same file states "The class percentages come from different and are approximate; the pinned total for TE-derived sequence is ~46%." Correctly scoped.
  2. part-01/06-rna-processing.md:266 — "how do ~19,000 genes build a human". This is a rhetorical quotation of the question, two lines after the exact figure 19,442 is stated with a citation. Not a competing claim. The several ~20,000 genes figures in ch 47, ch 56, qb-10 and ps-11 are multiple-testing denominators, where a round order-of- magnitude is the right thing to use.

6. Problem sets and labs

Problem sets — 11 present, structurally sound, but the count per set varies

11 of 11 present. Every set has a ## Where you went wrong section with a populated table, and every solution <details> block is balanced.

FileProblems<details> open/close"Where you went wrong"WYWW table rows
ps-01-molecular-foundations1010/10yes8
ps-02-mendelian-genetics88/8yes9
ps-03--and-mapping66/6yes8
ps-04--and-risk88/8yes16
ps-05--and-88/8yes14
ps-06-gene-regulation88/8yes15
ps-07-population-genetics88/8yes10
ps-08-quantitative-genetics88/8yes13
ps-09-molecular-evolution88/8yes15
ps-10-genomics-and-sequencing88/8yes14
ps-11-statistical-genomics88/8yes15

Nine of eleven have exactly 8 problems. ps-01 has 10 and ps-03 has 6. Each still has one worked solution per problem, so nothing is missing within a set — but if 8 is the intended invariant, two sets do not meet it. Flagged below rather than edited.

Labs — 8 present, 11 advertised

FAIL. README.md:8 and README.md:63 both promise "11 hands-on computational labs"; STUDY-GUIDE.md:7 says "11 labs". The directory contains 8:

labs/lab-00-setup.md               ## Check yourself  L225
labs/lab-01-sequences-and-fastq.md ## Check yourself  L205
labs/lab-02-alignment.md           ## Check yourself  L247
labs/lab-03-variant-calling.md     ## Check yourself  L249
labs/lab-05-assembly.md            ## Check yourself  L241
labs/lab-07-population-genetics.md ## Check yourself  L268
labs/lab-08-gwas.md                ## Check yourself  L226
labs/lab-10-phylogenetics.md       ## Check yourself  L217

Every lab that exists has a ## Check yourself section — 8/8 pass that check. The gaps are lab-04, lab-06 and lab-09, which the numbering leaves conspicuously empty. Nothing anywhere in the repo links to them, so this produced no broken links; the only symptom is the count mismatch against the README. See Needs attention.


7. Question banks

PASS. All 13 banks parse through reference/to_anki.py with exit code 0 and produce exactly 3 tab-separated columns on every row.

BankExitCardsRows with ≠ 3 columns
qb-part-000510
qb-part-010600
qb-part-020630
qb-part-030690
qb-part-040590
qb-part-050860
qb-part-060550
qb-part-070730
qb-part-080610
qb-part-090820
qb-part-100550
qb-part-110670
qb-part-120580
Total08390

Whole-directory run, confirming the same total:

$ python3 reference/to_anki.py question-banks/ --out all.tsv
  …
wrote 839 cards to all.tsv
rows: 839
rows with !=3 cols: 0

to_anki.py raises ParseError on a question with no answer or an answer with no preceding question, so exit 0 across all 13 banks also certifies that no Q: is orphaned and no A: is dangling.


8. Word count by section

SectionFilesWords
part-00-orientation26,808
part-01-molecular-foundations739,632
part-02-transmission-genetics738,925
part-03-genome-instability527,972
part-04-gene-regulation531,303
part-05-population-genetics423,238
part-06-quantitative-genetics316,077
part-07-molecular-evolution318,985
part-08-methods318,504
part-09-genomics745,875
part-10-functional-genomics536,810
part-11-human-and-statistical-genomics648,040
part-12-applications-and-ethics217,764
Chapters subtotal59369,933
problem-sets1145,831
question-banks1369,319
labs814,587
reference34,697
root (README, RESUME, STUDY-GUIDE)33,364
Total97507,731

Mean chapter length ≈ 6,270 words. Largest files: 58-ethics-and-society.md (9,306), 50-3d-genome.md (9,160), 54-rare-variants-and-mendelian-disease.md (9,149), 56-cancer-genomics.md (8,939).


Fixes applied

None. Nothing in the trivially-fixable category was found:


Needs attention

Three items, none auto-fixable, in descending order of how visible they are to a reader.

README.md:180 links to GLOSSARY.md from the Reference section. The file has never been written. RESUME.md corroborates rather than contradicts this: line 20 lists GLOSSARY.md as "in flight" and line 45 describes it as "in flight, 300–400 terms built from the chapters."

Two honest resolutions, and picking between them is an editorial call:

Silently repointing the link is not an option — the README's claim that every term is "defined once" is currently unbacked.

2. Three labs are missing against the advertised eleven

labs/ holds lab-00, 01, 02, 03, 05, 07, 08, 10 — eight files. README.md:8, README.md:63 and STUDY-GUIDE.md:7 all state eleven. lab-04, lab-06 and lab-09 do not exist.

The numbering gaps look deliberate rather than accidental — the labs that exist map onto chapter topics (alignment, , assembly, population genetics, , phylogenetics), and the three absent numbers sit exactly where an lab, an /data-formats lab and a functional-genomics lab would go. So this reads as "three labs still to write", not "three labs lost".

Either write them or correct the count in the two front-matter files. As it stands a reader counts the directory and finds the README overstated by three.

3. Problem-set length is not uniform

Nine sets have 8 problems; ps-01-molecular-foundations has 10 and ps-03-linkage-and-mapping has 6. Every problem in every set has a worked solution, so no set is internally incomplete — this is only a defect if 8 was meant to be an invariant. If it was, ps-03 is two problems short and ps-01 two over. Recording it rather than guessing at the intent.


Addendum — resolved after the report ran

This report is a point-in-time artifact and has deliberately not been rewritten. Recording what changed afterwards instead:

ItemStatus at report timeNow
1. GLOSSARY.md missingthe only broken link in the repoResolved. 397 terms, 14,398 words, 26 A–Z sections. Every entry links to the chapter that develops it; all 397 links resolve; all 59 chapters and all 13 parts are referenced. README.md:180 now resolves.
3. ps-03 had 6 problemsflagged as short of the apparent 8 invariantResolved. Two problems added (additive map building; sex differences and hotspot rate density), arithmetic independently recomputed. Now 8, <details> balanced 8/8.
2. Three labs missinglab-04, 06, 09 absent against 11 advertisedPartly resolved. lab-04 and lab-09 written and executed; lab-06 outstanding.

ps-01 retains 10 problems by intent — it is the foundational set and carries the extra strand/frame exercises the later sets build on. Eight is a norm, not an invariant.

One correction to §2 of this report. A re-run of the link check after GLOSSARY.md landed reports a single "broken" link at reference/verification-report.md:88 — which is this report quoting the - [Glossary](GLOSSARY.md) line as evidence, inside a fenced code block. The checker described below skips fenced blocks and does not produce this false positive; a simpler line-based checker does. Worth knowing if you rebuild the tool: skip code fences, or this report will report itself.

Reproducing this report

The link checker used here is not committed to the repo. Its logic, restated so it can be rebuilt: walk *.md excluding .venv/.git/.gstack, track fenced code blocks with a ```/~~~ toggle and skip their contents, match \[(?:[^\]]*)\]\(\s*([^)\s]+)(?:\s+"[^"]*")?\s*\) for inline links plus ^\s{0,3}\[[^\]]+\]:\s*(\S+) for reference definitions, skip targets matching ^(https?:|ftp:|mailto:|tel:|data:|#|//), split off # fragments, urllib.parse.unquote the rest, then os.path.exists(os.path.normpath(os.path.join(dirname(src), target))).

The other checks are one-liners:

# artifacts
grep -rn --include='*.md' -E '<invoke|<parameter name|antml:' . \
  --exclude-dir=.venv --exclude-dir=.git --exclude-dir=.gstack

# details balance (strip inline code first, or you get 3 false positives)
python3 -c "import re,sys; t=re.sub(r'\`[^\`\n]*\`','',open(sys.argv[1]).read()); \
  print(t.count('<details'), t.count('</details>'))" FILE.md

# trailing newline
find . -name '*.md' -not -path './.venv/*' -exec sh -c \
  'tail -c1 "$1" | read -r _ || echo "$1"' _ {} \;

# question banks
python3 reference/to_anki.py question-banks/ --out /tmp/all.tsv
awk -F'\t' 'NF!=3' /tmp/all.tsv | wc -l