A clinician asks about a mechanism roughly eighteen months before a paper about it appears. This is what the asking looks like from here: a profession, a country and a topic. No name, no employer, no initials, no avatar, and not one word the clinician typed. The link from a card back to a person stays inside the signal layer, on the far side of the wall DATA-02 draws, where no service in this product can read it.
SampleNothing below is a measurement.
Drawn, not measured. ADR 0002 puts aggregation and the release rules in a signal service that neither product has written, so this repository has no signal-event table, no aggregated-signal table and no route that reads one. OncoSkepsis has no deployment either: its first tumour boards run November to December 2026.
Check it: POST /api/signals/ingest on this deployment answers with that refusal, and GET /api/signals/boundary publishes the admissible-field allow-list, which is empty.
The four states of the contact model, side by side, because this build cannot perform the transition between them and four cards are honest where an animation would not be. A card carries no count of its own: a weekly figure for one person is that person’s activity, and no rounding makes a count of one safe to publish. The counts sit on the topic, below, where at least 8 different people stand behind every figure. OQ-05OQ-42
medical oncologist in Greece
sustained interest in KRAS G12C in lung adenocarcinoma
sustained topic signal
An interim reading (NP-15), not a rank and not a score.
Asking since 2026-W29, and still asking in 2026-W38. Two weeks, because the signal layer holds an ISO week and nothing finer. OQ-43
not contacted
Request contactA name is shared only if this person accepts, and only with the account that asked.
pathologist in Greece
sustained interest in KRAS G12C in lung adenocarcinoma
sustained topic signal
An interim reading (NP-15), not a rank and not a score.
Asking since 2026-W31, and still asking in 2026-W38. Two weeks, because the signal layer holds an ISO week and nothing finer. OQ-43
Kalypso Therapeutics asked in 2026-W38. No answer yet, and no part of this card changes while there is none. The record keeps the wording that was shown, word for word, and this card’s own page prints it. OQ-42
request sent
No name is shown while a request is outstanding. The request itself is the record, and it names the company that sent it.
medical oncologist in Ireland
sustained interest in HER2-low breast cancer
sustained topic signal
An interim reading (NP-15), not a rank and not a score.
Asking since 2026-W28, and still asking in 2026-W38. Two weeks, because the signal layer holds an ISO week and nothing finer. OQ-43
Kalypso Therapeutics asked in 2026-W36. Accepted in 2026-W37. The record keeps the wording that was shown, word for word, and this card’s own page prints it. OQ-42
accepted, name shared with this account
oncology pharmacist in Portugal
sustained interest in KRAS G12C in lung adenocarcinoma
sustained topic signal
An interim reading (NP-15), not a rank and not a score.
Asking since 2026-W33, and still asking in 2026-W38. Two weeks, because the signal layer holds an ISO week and nothing finer. OQ-43
Kalypso Therapeutics asked in 2026-W37. Declined in 2026-W38. The record keeps the wording that was shown, word for word, and this card’s own page prints it. OQ-42
declined
No contact detail, and no second ask from here. The decision is kept so that asking again would have to be a decision somebody takes on the record.
A topic is what the questions were about, coded in the vocabularies OQ-44 names. Two identifiers those vocabularies require are absent here rather than invented: this build has no service that resolves a ClinGen Allele Registry ID or a ChEMBL ID, and an identifier drawn to look right resolves to a real and different allele or molecule. Bars rather than a line, because a line drawn through a withheld week interpolates across it, and the interpolated point is a statement about a count the service refused to release. Every series is drawn against the same scale.
released at variant level as of 2026-W38
OncoTree LUAD Lung Adenocarcinoma, HGNC:6407 KRAS, NM_004985.5:c.34G>T p.(Gly12Cys), KRAS G12C inhibitor
Latest week 2026-W38: about 35 people
The gaps above, and the rules that made them:
Withheld: fewer than 8 different people stand behind this. 8 is the minimum, and there is no view, export, alert or API answer in this product that shows what is under it. OQ-14
Withheld: one institution supplies so large a share that a second could estimate its count. The dominance check runs at 5 percent and it can withhold a large number, so a gap here is not a small number. OQ-40NP-12
Which gap each rule took is not stated. Pairing them would say more about a single week than either rule allows.
The questions behind this line: tumour-board case, variant interpretation, country-specific literature check. OQ-45
released at biomarker level as of 2026-W38
OncoTree IDC Breast Invasive Ductal Carcinoma, HGNC:3430 ERBB2, HER2-directed antibody-drug conjugate
HER2-low: Immunohistochemistry 1+, or 2+ with a negative in situ hybridisation result. It is read from a stain rather than from a sequence, so there is no variant to code. OQ-44 names no code list for an expression category, so the definition is carried verbatim rather than coded to something it is not.
Latest week 2026-W38: about 25 people
The questions behind this line: tumour-board case, tumour-board brief, country-specific literature check. OQ-45
released at gene level as of 2026-W38
OncoTree IDC Breast Invasive Ductal Carcinoma, HGNC:3467 ESR1, selective oestrogen receptor degrader
Latest week 2026-W38: about 25 people
The questions behind this line: variant interpretation, tumour-board case, reimbursement dossier. OQ-45
released at variant level as of 2026-W38
OncoTree PAAD Pancreatic Adenocarcinoma, HGNC:6407 KRAS, NM_004985.5:c.34G>T p.(Gly12Cys), KRAS G12C inhibitor
Withheld: fewer than 8 different people stand behind this. 8 is the minimum, and there is no view, export, alert or API answer in this product that shows what is under it. OQ-14
released at variant level as of 2026-W38
OncoTree COADREAD Colorectal Adenocarcinoma, HGNC:6407 KRAS, NM_004985.5:c.34G>T p.(Gly12Cys), KRAS G12C inhibitor with an anti-EGFR antibody
Withheld: fewer than 3 different institutions stand behind this. 3 is the minimum, and one or two institutions asking about a topic is close enough to naming them. OQ-40
What the rules do to a variant too few people asked about: it is not dropped and it is not hidden, it is counted in its gene and the gene row says so. OQ-41
| Gene | Week | Asked about by | Variants broken out |
|---|---|---|---|
KRAS HGNC:6407Lung Adenocarcinoma (LUAD) | 2026-W38 | about 35 people | G12C about 25 people NM_004985.5:c.34G>T p.(Gly12Cys) |
ESR1 HGNC:3467Breast Invasive Ductal Carcinoma (IDC) | 2026-W38 | about 25 people | D538G about 15 people NM_000125.4:c.1613A>G p.(Asp538Gly) |
ERBB2 HGNC:3430Breast Invasive Ductal Carcinoma (IDC) | 2026-W38 | about 25 people | None. |
KRAS The other KRAS variants asked about this week are counted in the gene row and are not broken out, because each of them on its own is below the minimums. There is no control on this screen that would show them, and a larger account does not get one. OQ-14
ESR1 The remaining ESR1 variants are counted in the gene row. They are not named here: naming a variant that rolled up would publish that it had activity below the minimum, which is the thing the roll-up withheld. OQ-14
ERBB2 Nothing rolled up into this row.
A variant appears at variant level only when it meets the same two minimums the gene row met. Otherwise its questions are counted in the gene row and are not broken out, which is why a gene total is larger than the variants under it and why the difference between the two is not a number this screen can give you. OQ-41OQ-14OQ-40
| Topic | Earlier | Later | Change |
|---|---|---|---|
| KRAS G12C in lung adenocarcinoma | 2026-W29 to 2026-W33 about 20 people | 2026-W34 to 2026-W38 about 45 people | +25 people |
| HER2-low breast cancer | 2026-W30 to 2026-W34 about 25 people | 2026-W35 to 2026-W38 about 35 people | +10 people |
| ESR1 variant level, breast invasive ductal carcinoma | 2026-W29 to 2026-W33 withheld | 2026-W34 to 2026-W38 withheld | withheld |
ESR1 variant level, breast invasive ductal carcinoma Withheld: comparing these two periods could recover a count the rules already withheld. Both periods go, not only the difference, because a difference with one side published is one subtraction away from the other. NP-10X-11
A period figure counts the different people across the whole period, so it is larger than any single week in it and smaller than the weeks added together. Adding the bars above will not reproduce it, and neither will dividing one of these rows by another.
Export refused Send to CRM refused
Neither control is wired to anything, and neither is a spinner: this API has no export route at all, and FR-INT names the integrations nobody has built. If one is written, the differencing check runs before it answers and refuses the output rather than the view, which is why the withheld row above is on the screen and an export of it would not be. NP-10X-11XFER-10
Ten. A figure appears only when at least 8 different people and 3 different institutions stand behind it, and every figure is rounded to the nearest 5, so 8 and 9 both display as 10 and nothing on this deployment can show you a 1, a 3 or a 7. The rounding is applied in every view, every export, every alert and every API answer, and the figure it was rounded from is not sent to this screen at all.
The dominance check is the part that surprises people: a week can be withheld because one hospital supplied most of it, however large the total was. A gap in a series is therefore not a small number, and from here the two cannot be told apart, which is the point rather than a limitation of the screen.
The wall that enforces all of this runs live. Hand it a payload with a medical record number in it and it names the field, the rule and the requirement that refused it. Nothing on that screen is drawn.
Absent by decision