Imagine a pot of water on a stove. The burner is set somewhere on the dial — low, medium-low, medium, medium-high, high — now imagine you set the dial to whatever temperature you choose. That determines the temperature your water is going to be. If the burner stays too high for too long, the water reaches a boil. No one is around to see it starts spilling over the rim. Left that way unattended, the water boils away entirely. Nothing you can do now will ever get that water back into the pot. That pot is a human cell population. The burner is whatever has been driving that cell population to do more cellular work than it should — chronic inflammation, chronic infection, carcinogen exposure, autoimmune dysregulation, prolonged hormonal exposure. The water reaching a boil is the moment a tumor crystallizes and starts producing the symptoms that send a patient to a clinic. Unfortunately today, the only thing that can be detected is steam — the water must be at a raging boil to be detected by anything that currently exists. This framework changes that. It reads both the temperature of the water and the setting of the dial. It reads where the cell population is currently sitting on the temperature scale, calibrated against thermodynamic floors derived from physics, and it reads in both directions.
📅
10+ Year Pre-Diagnostic Window — Breast Cancer
Pink ribbon · Secretory class · VAL-047 · EPIC-Italy GSE51057
The EPIC-Italy cohort is a real-world prospective study: women donated blood at study enrollment with no idea whether they would later develop cancer. Years later, some did. When the framework reads the blood drawn more than ten years before those women would walk into an exam room with a breast cancer diagnosis, the immune-class architectural signature is already loud. The reading climbs monotonically with the pre-diagnostic interval — quiet near diagnosis, louder at five years, louder still at ten — and the longest-window effect size is among the strongest pre-diagnostic signals in the methylation literature. The result replicates independently in a second EPIC-Italy sub-cohort drawn on the same 450K platform with the same panel. A woman sitting in a clinic today, told she is fine, may already be carrying a readable cellular signature of a disease that will not declare itself for ten years. We have been telling people they are fine when their own bodies have been trying to say otherwise for a long time.
Sources: Severi 2014 (EPIC-Italy GSE51057, GSE51032) · Xu 2020 JNCI (Sister Study panel) · VAL-093 + VAL-096 (cookbook).
🌐
Multi-Cancer Systemic Drift
VAL-046 · Pancreatic, Lung, Colorectal, Prostate · 4 independent cohorts
Look at the same EPIC-Italy blood from a decade before diagnosis — but now from the per-tissue angle, asking which cells the drift is coming from. The answer is not what most people would guess. At ten-plus years before clinical breast cancer, the loudest tissue tiles are pancreatic, kidney, head-and-neck, and colon. The breast tile reads near-null at this window. The cellular system as a whole is showing wear earlier than it should for the customer's age, and a specific tissue tile lights up only in the months before clinical diagnosis. Both components replicate concordantly across two independent cohorts. The pre-diagnostic signal is not "early breast cancer" — it is a body that is already aging differently than its calendar age would predict, with the localized cancer arriving on top of that body-wide drift years later.
Sources: VAL-093 + VAL-096 (cookbook) · Loyfer 2023 array atlas (Stage 2 cell-of-origin reference) · Severi 2014 (EPIC-Italy).
📏
Organ-Wide Field Effect
VAL-037 + VAL-039 · 24 TCGA cancer types · n = 1,109 STN
"Adjacent normal" tissue in a pathology report is the surrounding tissue a surgeon takes as a clean control alongside the tumor itself. It is supposed to be healthy. The framework reads it differently. Across two dozen TCGA cancer types, adjacent-normal tissue sits noticeably above the reference of cells that have never been near a tumor — about a quarter of the way from healthy toward the tumor reading. Six studies that recorded distance from the tumor show a clean gradient: closer to the tumor, more drift; further away, less. Even five to ten centimeters out, the drift is still present. What clinicians have been calling a clean margin is, by this reading, an organ-wide architectural change that standard histopathology was never built to see. Whether that has implications for recurrence rates or for surgical margin guidelines is an open clinical question that pathologists, not the framework, will need to answer.
Sources: VAL-037 + VAL-039 (cookbook) · TCGA pan-cancer methylation reference · six distance-annotated tissue studies (lung, breast, colon, prostate, HCC, gastric).
Colorectal — Full Trajectory
Dark Blue ribbon · VAL-042 + VAL-044 + VAL-046
The colon adenoma-to-carcinoma sequence is the best-documented cancer progression in human medicine. The framework reads each clinical stage as a different reading on the same gauge — healthy colon, hyperplastic polyp, tubular adenoma, advanced adenoma with high-grade dysplasia, invasive cancer — and the gauge climbs monotonically through all five. Tubular adenoma crosses into the framework's pre-cancer tier; invasive disease crosses the floor. When a CRC patient goes through FOLFOX chemotherapy, responders show the gauge dropping back toward the healthy band and complete responders return to normal. Non-responders stay elevated. The framework also reads the tumor itself across two separate cell populations inside a single biopsy — the tumor cells against their own architectural floor, and the immune cells that have moved into the tumor against the immune-class panel. Colonoscopy raised colorectal cancer survival from fifty percent to sixty-five percent over four decades. It did not cure cancer. It moved the diagnosis earlier, when treatment still works. A blood-based cellular-state reading offers the same kind of opportunity, applied to people in the years before they become eligible for routine colonoscopy at all.
Sources: Hou 2012 Am J Epidemiol · Luo 2014 Gastroenterology · Parikh 2019 Nat Med · Reinert 2019 JAMA Oncol · VAL-061 + VAL-062 (TCGA-COAD HM450 paired tumor/adjacent-normal). ACS Colorectal Cancer Facts & Figures 2023-2025 (survival trend).
Breast — A Decade of Runway
Pink ribbon · Secretory class · VAL-047 + VAL-046 + VAL-044 + VAL-039
Mammography is the most successful cancer screening program in the history of American medicine. Five-year breast cancer survival rose from seventy-five percent in the mid-1970s to ninety-one percent today, and the change came not from curing cancer differently but from finding it earlier. The framework asks what additional runway is possible if the cellular signal can be read in blood years before any imaging modality could see the tumor.
The answer in two independent EPIC-Italy cohorts: a cellular signature is loudly present more than ten years before clinical diagnosis, climbs monotonically with the pre-diagnostic interval, and replicates concordantly across both cohorts. By the two years before diagnosis, the breast tissue tile itself starts to localize. About half of women who go on to develop breast cancer are already sitting in the upper decile of their age-matched healthy peers a full decade before their diagnosis, and the fraction climbs further as the diagnostic window approaches.
The same instrument addresses the dense-breast problem mammography cannot. Forty percent of American women have dense breast tissue, and mammographic sensitivity drops sharply in that population. The framework reads methylation entropy, not X-ray absorption — dense and non-dense tissue share the same architectural floor.
The screening did not cure cancer. It moved the diagnosis earlier, when treatment still works. The pre-diagnostic signal documented here is the cellular substrate of that same opportunity, with a decade of runway instead of two years, on a blood draw most patients are already getting for other reasons.
Sources: Severi 2014 (EPIC-Italy GSE51057, GSE51032) · Xu 2020 JNCI (Sister Study panel) · ACS Breast Cancer Facts & Figures 2024-2025 (survival trend) · VAL-093 + VAL-096 + VAL-060 (cookbook).
Pancreatic — The Largest Near-Term Runway
Purple ribbon · Secretory class · VAL-046 + VAL-007 + VAL-040
Pancreatic adenocarcinoma carries the lowest survival rate of any major cancer in the United States. Across all stages combined, five-year survival sits at thirteen percent. For patients diagnosed with metastatic disease — which is most patients at presentation, because the disease is typically silent until it has spread — five-year survival drops to three percent. The American Cancer Society projects approximately sixty-eight thousand new cases in the United States in 2026 with approximately fifty-three thousand deaths.
There is no recommended general-population screening test. The disease is usually caught when a patient develops jaundice, weight loss, or new-onset diabetes severe enough to trigger an imaging workup, by which point the window for curative surgery has typically closed. Stage IA pancreatic cancer caught in time for the Whipple procedure has five-year survival above thirty percent in real-world cohorts — a tenfold improvement over the all-stages baseline.
The framework's pancreatic application reads a cellular signature in pre-diagnostic blood drawn years before the cancer presents. The framework also flags a clinical pattern oncologists already recognize but currently have no instrumental screening for: new-onset Type 2 diabetes after age fifty paired with any directional cellular-departure signature is a paraneoplastic-PDAC workup trigger.
If even a fraction of pancreatic cancers were caught at the stage where the Whipple is curative instead of palliative, the change in absolute survival would be measured in tens of thousands of lives per year in the United States alone.
Sources: SEER + ACS 2026 mortality projections · VAL-046 (Rotterdam Study pre-diagnostic) · VAL-069 (TCGA-PAAD 324-CpG directional fallback panel) · VAL-007 (pancreatic exocrine cfDNA tissue-specific signal).
Glioma — LGG & GBM
Grey ribbon · Terminal class (tightest floor) · VAL-007 + VAL-044
Brain cancer has historically been undetectable in standard peripheral blood at array resolution. Cell-free DNA from neurons appears in the bloodstream at fractions so small that conventional methylation reference atlases were calibrated assuming the signal would not be reachable. The framework's earlier framing accepted this as a structural limitation — read brain disease in cerebrospinal fluid, not plasma.
That framing was wrong, and the reason is mundane: the older atlases used a bulk-tissue brain reference that did not separate cortical-neuron signal at the resolution standard methylation arrays measure. When a more recent atlas with a sorted-cell cortical-neuron entry is used as the Stage 2 reference, the signal in glioma plasma reads cleanly. Glioma plasma carries roughly four times more cortical-neuron cell-free DNA than healthy reference plasma, and the same effect appears in pre-surgery treatment-naive samples — meaning the signal is the disease, not the surgery or chemotherapy.
The same methodology run against Alzheimer's blood reads cortical-neuron cfDNA at near-null. AD does not elevate cortical-neuron cfDNA at this resolution, while glioma does. For a patient presenting with cognitive symptoms, the same instrument reads whether the cellular signature is consistent with neurodegeneration or with a tumor — from a single peripheral blood draw, with no need for a lumbar puncture, contrast MRI, or amyloid PET to start the differential.
Sources: VAL-090 (GSE180683 EPIC peripheral blood, n=76 glioma vs n=177 healthy) · VAL-091 (AD layered-atlas Stage 2 null) · Loyfer 2023 array atlas · Salas-Wiencke 2022.
Alzheimer's — Multi-Class at the Thermodynamic Level
Purple ribbon · Multi-class (4 elevated) · VAL-040
Alzheimer's disease has approximately 6.9 million prevalent cases in the United States today and is projected to roughly double by 2050 if no preventive intervention emerges. The new amyloid-targeting therapies — lecanemab and donanemab — are FDA-approved only for patients in mild cognitive impairment or mild dementia stages, with confirmed elevated brain amyloid. The drugs are not effective once the patient has progressed past mild dementia. The clinical bottleneck is identifying patients early enough that disease modification can do meaningful work, and current diagnostic pathways rely on cognitive testing that detects the disease only after symptoms have begun.
The framework reads AD as a systemic multi-class phenomenon at the cellular thermodynamic level — not a localized brain event. Four cell architectural classes show drift in AD cohorts: brain cortex itself, peripheral blood immune cells, pancreatic islets, and cerebral vasculature. Every tested tissue-class combination shows a severity gradient from early to late disease. The directional panel built for AD blood-based screening reads a real signal years before clinical presentation in independent cohorts.
A client identified as showing AD-pattern drift could enter monitoring earlier, begin pharmaceutical intervention earlier when those interventions still work, and have time to make the legal, financial, and family decisions that AD eventually forces. For some, that lead time may be the difference between writing the will yourself and having someone else write it for you. For some it may be the difference between being there for a grandchild's first communion and not being there.
Sources: De Jager 2014 Nat Neurosci (ROSMAP n=708) · Shireby 2022 Brain (BDR n=1,408) · Nabais 2021 Genome Biol (peripheral blood meta-analysis n=3,424) · Lunnon 2014 Nat Neurosci · VAL-040 + VAL-051 (cookbook).
🧭
Tissue-of-Origin Localization
VAL-041 · 10 cancer types · 10/10 top-1 correct
A standard plasma sample contains cell-free DNA fragments from every tissue in the body simultaneously, mixed together. Reading the bulk plasma against any single tissue's architectural floor is a category error — the math does not work because the signal is averaged across cells the architecture is not for. The fix is to first separate the bulk into per-tissue fractions using a published methylation atlas, then score each separated fraction against the architectural floor of its own class. When this is done correctly, the framework localizes the tissue of origin correctly in every cancer type tested. The current implementation runs a layered atlas — the original Moss 2018 reference, plus the more recent Loyfer 2023 array atlas that supplies the cortical-neuron resolution that opened brain cancer in plasma. Stage 2 of the pipeline is what turns a routine annual blood draw into a tissue-resolved cellular-state reading, on the same sample most patients are already giving for cholesterol and metabolic panels.
Sources: Moss 2018 Nat Commun · Loyfer 2023 Nature (array atlas) · Liu 2020 Ann Oncol (extended marker panel) · VAL-041 (cookbook).
Seminoma — The Inversion Signature
Orchid ribbon · Pluripotent class · VAL-045
Most cancers read on the framework's gauge as elevation above the architectural floor — a cell is doing more methylation work than it should be. Seminoma and embryonal carcinoma read in the opposite direction. The pluripotent cells these tumors arise from sit, in their healthy state, near the upper limit of methylation entropy that any somatic cell can reach. There is nowhere up to go. When these cells transform, the methylation pattern moves down toward hypomethylation, away from the floor in the opposite direction from every other cancer the framework has been calibrated against. A detector built only to flag elevation would miss the entire signal — and roughly five thousand testicular germ cell tumor cases occur annually in the United States, the seminoma fraction making up about sixty percent of them. Reading in both directions by construction is the difference between catching those five thousand patients and missing them.
Sources: Shen 2018 Cell · Killian 2016 Cell Rep (seminoma hypomethylation) · VAL-045 (cookbook).
🐕
Cross-Species Validation — Canine Oncology
VAL-043 · 5 canine cancers · n = 104 Labradors
If the framework is reading a real physics quantity rather than a human-cohort artifact, the same architectural floors should appear in non-human mammals. The test ran five canine cancers against the same class floors calibrated on human tissue, and the cross-species offset between dog and human readings is essentially zero across seventy million years of evolutionary divergence. Dog aging tracks human aging on the same gauge with near-perfect correlation. Dogs age roughly seven times faster than people, which means a prospective canine cohort can deliver a longitudinal validation in eighteen months that a human cohort would take a decade to produce. Veterinary oncologists and the dogs in their clinics are how the next layer of evidence becomes available before the human cohorts catch up — and in the process, the same instrument helps the dogs.
Sources: Wang 2020 Cell Reports (Labradors) · Pal 2016 Cancer Res · Beck 2020 Vet Comp Oncol · Decker 2015 PLoS Genet · VAL-043 (cookbook).
📉
Treatment Response Trajectory
VAL-044 · 5/5 clinical trials · GBM, CRC, BRCA, AML, melanoma
The same gauge that climbs as a cell population drifts away from healthy can drop again when treatment works. Across five tested clinical trial cohorts — glioblastoma on the Stupp protocol, colorectal cancer on FOLFOX, breast cancer in adjuvant chemotherapy, acute myeloid leukemia on 7+3 induction, and melanoma on anti-PD-1 — the framework's reading separates patients who respond to therapy from patients who do not. Responders show the gauge dropping back toward the healthy band. Complete responders return to normal. Non-responders stay elevated or continue to rise, and the divergence is visible in serial samples drawn during treatment.
Acute myeloid leukemia shows the strongest single-cohort signal in the framework's catalog. AML cancer cells are the same myeloid-lineage cells the universal Stage 1 panel was built to read, which means the framework reads the disease cells directly rather than against an immune-cell background. The signal at diagnosis is essentially impossible to miss.
Architectural drift precedes the disease. Architectural recovery accompanies the treatment that works. Both are measurable in the same blood draw — which means a patient on therapy, and the oncologist managing that therapy, can see whether the treatment is working months before standard imaging would show it.
Sources: Ceccarelli 2016 Cell + Fan 2021 J Neurooncol (GBM) · Parikh 2019 Nat Med (CRC) · Stover 2018 J Clin Oncol (BRCA) · Ley 2010 NEJM + Ding 2012 Nature (AML) · Cabel 2018 Ann Oncol (melanoma) · Glass 2017 (GSE62298 AML) · VAL-044 + VAL-082 (cookbook).
Prostate — PIN to Metastatic
Light Blue ribbon · Secretory class · VAL-042 + VAL-046
Prostate cancer screening has been arguing with itself for thirty years. PSA testing flags many men whose biopsies turn up nothing dangerous, and the concern about overtreatment has shifted clinical guidelines back and forth more than once. The clinical question is not whether a screening test can detect — it is whether a screening test can stratify, telling clinicians which men in the gray zone actually need workup and which can safely watch and wait.
The framework reads the full clinical progression — healthy prostate, low-grade prostatic intraepithelial neoplasia, high-grade PIN, Gleason 3+3, Gleason 4+3 and higher, metastatic castrate-resistant — as six distinct readings on the same gauge, each higher than the last. The gauge crosses into the pre-cancer band at HGPIN and breaches the floor at clinically significant disease. In a separate prospective cohort, blood drawn three years before prostate cancer diagnosis already shows immune-class drift. Prostate-epic was also the first card in the framework's catalog built tissue-first, on a published cohort of African-American men where direct tissue measurement carries clear signal.
Roughly one in five men who undergo radical prostatectomy develop long-term urinary incontinence, and roughly two in three experience long-term erectile dysfunction. The clinical cost of treating the wrong prostate cancer is high. A reading that helps the urologist decide which man in the PSA gray zone actually needs the biopsy — and which one can watch and wait another year — is the kind of information the current toolkit does not have.
Sources: Jerónimo 2008 Clin Cancer Res · Aryee 2013 Sci Transl Med · Damaschke 2017 (zonal gradient) · Horvath 2014 Health ABC · Berglund/Yamoah/Kresovich 2024 (GSE269244, n=238 African-American men, EPIC 850K) · VAL-042 + VAL-046 + VAL-058 (cookbook).
Complete Validation Record — VAL-001 through VAL-128
128+sealed validation studies executed
24+cancer types tested across the cascade
4,304matched tumor-normal pairs (TCGA pan-cancer)
30TCGA cancer types · 29/30 confirmed
10+ yrbreast pre-diagnostic window (VAL-047)
0free parameters · zero cancer training data
The full testing record at a glance. The methylation core (VAL-001 to VAL-013) established the pan-cancer A-score on TCGA, Health ABC, Roadmap Epigenomics, Moss 2018 cfDNA atlas, OSK reprogramming, DunedinPACE, pan-mammalian aging, and Labrador cross-species. The multimodal expansion (VAL-014 to VAL-033) added the five-substrate framework across nucleosome occupancy, fuzziness, WPS, and fragment-size alongside methylation. The vertebrate extension (VAL-034 to VAL-036) confirmed pan-mammalian lifespan correlation and ectotherm temperature correction. The multi-class drift cascade (VAL-037 to VAL-046) added organ-wide field effect, tissue-of-origin localization, pre-cancer progression, treatment response, inversion specificity, cross-species replication, multi-class AD drift, and the pre-diagnostic multi-cohort signature. VAL-047 closed the cascade loop with external per-patient 450K methylation on 1,581 real samples from three GEO cohorts.
The per-card tissue and pattern wave (VAL-048 to VAL-100) added card-specific tumor-vs-adjacent-normal validation on public TCGA cohorts — moving breast, lung, hcc, prostate, and crc to tissue_arm_validated tier and surfacing CRC's blood-vs-tumor-TIL compartment-direction-flip. The pancreatic-epic 324-CpG directional fallback panel established the pooled-null + directional-pass pattern as the second case alongside AD (TCGA-PAAD d = +1.51, 7/7 patients positive). Cervical-epic v0.1 documented the first opposite-sign Cohen's d across independent cohorts of the same disease, creating the exploratory_with_cohort_heterogeneity tier. Layered-atlas Stage 2 demonstrated glioma plasma reads cortical-neuron cfDNA at d = +1.96 — overturning the earlier prediction that plasma fails for terminal class — and confirmed AD does not elevate cortical-neuron cfDNA at array-NNLS resolution, enabling glioma-vs-AD differential diagnosis from a single IDAT. The first multi-cohort run-everything pass surfaced replicating sub-cell-type resolution gains in breast pre-diagnostic blood (UniLIFE 19-cell aTreg at >10yr, aBnv at 0-2yr).
The multi-atlas calibration discipline wave (VAL-101 to VAL-128) anchored every active card on a structurally-separated TCGA n=210 healthy calibration cohort with per-atlas q5 thresholds sealed before any disease scoring. Cardio-epic v0.3 (VAL-108-113), prostate-epic v0.3 with the ProstateRef luminal dedifferentiation pattern at d_paired = −0.767 (VAL-117-118), bladder-epic v0.1 with the cohort-substrate-coverage gate (VAL-119-122), and gastric-esophageal-epic v0.1 with within-cancer histological-subtype discrimination at d_ESCC−EAC = −1.06 (VAL-123-128). The wave produced 10 new cross-card lessons and 13+ new pre-flight checklist gates. All 128+ sealed VAL scripts and their JSON pass/fail records are in the public repository. Pre-clinical research only; prospective clinical validation required before individual-patient use.