Luminary Studies · environment module 01
Four compartments, because the lumen, an intact crypt, an invasive tumor and a liver metastasis are four different places that happen to share a name. Two tracks: the disease left alone, and the disease treated. Every component traces to a published measurement, and every timing figure is marked for what it is — nobody has watched this happen; all of it is inferred from mutation counts and screening statistics.
11 October 2026 · nothing here was measured by us · measured · inferred · asserted are marked separately throughout
—
—
With the measurement that puts them in the drawing.
Printed, not filled in.
Directional edges, each one something that can be tested. This is the part a parts list cannot give you: the components are only interesting because of what they do to each other.
| from | → to | effect, with the number | where |
|---|
One founder cell to death, with every interval marked for how it was obtained. The solid bars were counted. inferred intervals were computed from mutation counts divided by an assumed mutation rate and an assumed cell-division time, or back-calculated from screening prevalence — change an assumption and the number moves.
| interval | figure | how it was obtained | source |
|---|---|---|---|
| Founder cell appears | before age 20 | methylation-drift clock, 675 samples inferred | Luebeck 2018 |
| Whole premalignant period | ~60 years mean | same clock, 322 TCGA samples inferred | Luebeck 2018 |
| Crypt fission, normal colon | 0.028 / crypt / year | 136,899 crypts scored in 20 patients — one crypt cycle every 36 years measured | Baker 2014 |
| Non-advanced → advanced adenoma | 3.3% / year at 55–59 | back-calculated from screening-colonoscopy prevalence against registry incidence inferred | COSIMO / Brenner |
| Advanced adenoma → preclinical cancer | 2.6% / yr at 55–59 · 5.1% at 70–74 | same model; the hazard roughly doubles with age inferred | COSIMO / Brenner |
| Large adenoma → advanced carcinoma | 17 years (CI 10.9–26.5) | 13 patients, comparative lesion sequencing; depends on an assumed 4.6×10⁻¹⁰ /bp mutation rate and a 4-day division time inferred | Jones 2008 |
| Carcinoma → metastasis | 1.8 years (CI 0.9–3.1) | 10 paired primary/metastasis, 233 mutations inferred | Jones 2008 |
| Preclinical → clinically apparent | 15.5–18.2% / year | screening-calibrated inferred | COSIMO |
| Median time to recurrence after resection | 13–15 months | 5,412 resected stage I–III, 877 recurrences measured | Neth. Cancer Registry |
| 5-year survival by stage | I 93.2% · IIA 84.7% · IIB 72.2% · IIIA 83.4% · IIIB 64.1% · IIIC 44.3% · IV 8.1% | 119,363 patients measured. IIIA above IIB is a real feature of that dataset, not a typo | O'Connell 2004 |
| Metastatic disease, supportive care only | 5.0 months median | the control arm of one randomized trial, n ≈ 13 — the only genuine untreated-survival figure we could source measured | Scheithauer 1993 |
Every interval in this disease's timeline is inferred. The spine of the natural history — 17 years to become a cancer, 1.8 years to spread — rests on an assumed mutation rate and an assumed division time, and is the least directly observed thing on the page.
Colorectal module, first pass. Nothing here was measured by us and no molecule described is synthesized. Microvessel densities are labeled with their counting method — hotspot and random-field counts differ systematically and are never placed on a shared axis. Where a source contradicted itself the contradiction is recorded rather than resolved by choosing the convenient value.