The programme
Drugs fail in target-positive patients because presence is not usability.
A target is counted, not tested. Counting confirms the label is there. It cannot say whether the population under that label still works, or still comes back after a dose. That is the difference between a patient who responds and one who does not, and a large share of late-stage attrition sits in it.
01 · The failure that has no name
A patient can score positive on every test and have almost none that still functions.
A receptor is called a target because it can be counted. Transcript, total protein, occupancy and diagnosis all report the same thing: the label is there. None of them reports whether the population under that label can still read a signal, pass it on, and come back for the next dose.
An enzyme cuts the binding site off a receptor and the assay still counts it. A receptor is made and never reaches the surface. A receptor binds and the machinery behind it is gone. Each of these produces a target-positive patient who does not respond, and each is a different problem requiring a different repair. Today they are one undifferentiated failure called non-response.
Non-response stops being an unexplained outcome and becomes a lesion map.
| What is observed | What is actually happening |
|---|---|
| Target positive, drug binds, no response | Usable target capacity is absent despite presence |
| Biomarker normal, patient deteriorating | Compensation preserves output at rising cost |
| Treatment works, then stops working | Return, reconstruction or closure of the target failed |
| Relapse after apparent recovery | A source, a retained state or a selection pressure persisted |
| One insult, multisystem picture | A high-gain interface routed a local defect into systemic output |
Stratify by the failed operation, not by diagnosis alone.
02 · What the programme builds
One engine. Five surfaces it presents to the outside.
These are not five businesses. They are one causal architecture, read at five points. The architecture is what makes each of them possible, and it is also what makes them consistent with each other.
Target-capacity pharmacology
A therapeutic category defined by the operation a target needs rather than by the disease name. A cut receptor calls for one operation. A receptor that cannot return after use calls for another. The measurement picks the operation, and the operation picks the molecule.
State-resolved companion diagnostics
Assays that report which failure mode a patient's target population is in, before the prescription rather than after it. The output is a decision, not a score.
Dynamic reserve measurement
Measurement of what a system can deploy, recover and use again, rather than what is present at rest. This is the standard the law implies, and the one no current test provides.
Human receiver and interface atlases
Adult human, donor-resolved maps of the blood-facing interfaces and of receiver composition across them. This is the substrate the rest of the engine reads, and it is built from human tissue rather than translated from models.
Treatment-sequence intelligence
Which operation runs first, what has to be rebuilt before the next agent can work, and when a support can be withdrawn to prove restoration rather than to assume it.
The scarce asset is the mapping
Not the atlas, and not any single molecule. The mapping from a measured failure to the operation that corrects it is what nobody else has, because nobody else has the causal object it is derived from.
Measure the limiting state, correct the limiting operation, prove that the system can do it again.
03 · Evidence that the approach finds targets
The architecture located the target. The drug worked. The trial was somebody else's.
GDF15 is a tissue-stress signal from the input layer. Its receptor sits almost entirely at one blood-facing hindbrain site, which is exactly the kind of arrangement the sentinel architecture predicts should carry systemic consequence out of proportion to its size.
In a randomised phase 2 trial, 187 patients with cancer cachexia received a monoclonal antibody against that signal or placebo. Every dose beat placebo on weight at 12 weeks, and the highest dose also improved appetite and physical activity.1 The chain runs end to end: a signal in the blood, a receptor at one interface, a drug that works.
We did not run that trial and we do not claim it. We point at it because it is the completed instance of the search pattern, and because it is the reason to believe the pattern will find the next one.
The law. In cancer cachexia the body meets every demand the tumour creates and never rebuilds the muscle it spent. Muscle loss is detectable on scans up to 18 months before diagnosis,2 and after diagnosis the patients losing muscle fastest carry roughly twice the risk of death, while fat loss over the same window carries none.3 The harm tracks the missing rebuild rather than the demands.
The receiver. The receiver discovery predicts that abundance and usable competence come apart. Postmortem series in sudden infant death syndrome find roughly twice the count of serotonin neurons alongside lower receptor binding,4 with the transmitter and its synthesising enzyme both reduced,5 in the nuclei that run the failed rescue. The direction was fixed before the data were opened. A later high-risk series then reversed the receptor direction in preterm infants, higher binding rather than lower, and still landed on the same failure.6
Full detail on the science page.
Three predictions, three completed tests, none of them on data we generated.
04 · Where this actually stands
The rung climbed, and the next one.
Positioning is earned in sequence. We state the rung reached and the rung targeted, and we do not skip ahead. Anyone can hold us to this list.
No claim on this site depends on a rung we have not climbed. The diagnostic logic and the measurement standards are entering validation. They are not validated practice, and we will not describe them as such until they are.
05 · What is open, and what is held
The class discoveries are public. The instruments are not.
This is a deliberate split, and it is the reason this site describes measurements without specifying them.
The law, the architecture and the failure classes belong in the literature, because a causal object that nobody can check is not a scientific finding. They are published, and the falsifiers are published with them.
The instruments are held: the assays that read a receiver's failure mode, the rules that select one repair over another, the panel compositions, the thresholds, and the protocols that verify restoration. These are filed rather than described, because the value of the programme sits in the mapping from a measured failure to the correcting operation, and that mapping is the part a competitor cannot rederive from the architecture alone.
The science can be evaluated before any agreement, because it is written to be reproduced.
The instruments are shown under agreement, in the order that fits the discussion: a single receptor programme, a disease-specific stratification, or the measurement standard itself.
A patent portfolio covering target-capacity therapeutics, state-resolved diagnostics, dynamic reserve measurement, developmental reserve and external-effector systems is in preparation. Filing status is discussed under agreement rather than published here.
Judge the architecture in public. Judge the instruments in the room.
References
The evidence cited on this page.
Every result above was produced by another group and published. None of it depends on data we hold.
- Groarke JD, Crawford J, Collins SM, et al. Ponsegromab for the treatment of cancer cachexia. New England Journal of Medicine. 2024;391(24):2291-2303. doi:10.1056/NEJMoa2409515 ↩
- Babic A, Rosenthal MH, Sundaresan TK, et al. Adipose tissue and skeletal muscle wasting precede clinical diagnosis of pancreatic cancer. Nature Communications. 2023;14(1):4317. doi:10.1038/s41467-023-40024-3 ↩
- Babic A, Rosenthal MH, Bamlet WR, et al. Postdiagnosis loss of skeletal muscle, but not adipose tissue, is associated with shorter survival of patients with advanced pancreatic cancer. Cancer Epidemiology, Biomarkers and Prevention. 2019;28(12):2062-2069. doi:10.1158/1055-9965.EPI-19-0370 ↩
- Paterson DS, Trachtenberg FL, Thompson EG, et al. Multiple serotonergic brainstem abnormalities in sudden infant death syndrome. JAMA. 2006;296(17):2124-2132. doi:10.1001/jama.296.17.2124 ↩
- Duncan JR, Paterson DS, Hoffman JM, et al. Brainstem serotonergic deficiency in sudden infant death syndrome. JAMA. 2010;303(5):430-437. doi:10.1001/jama.2010.45 ↩
- Kinney HC, Folkerth RD, Nelson ME, et al. Serotonergic receptor binding in the brainstem in the sudden infant death syndrome in a high-risk population. PLOS ONE. 2025;20(9):e0330940. doi:10.1371/journal.pone.0330940 ↩
Contact
The conversations worth having.
Three of them, in the order they usually happen.
A target that failed in target-positive patients
Bring the programme. The question is which failure mode the target population is in, and whether it is correctable.
A trial that needs a stratification
The measurement runs before randomisation, and the second response is what proves restoration.
The programme itself
For investors: what is built, what is filed, what the next result has to show, and when.