Assessment — idea-003

A native-plasma ctDNA (circulating tumour DNA) quality control material, sold under 21 CFR (Title 21 of the Code of Federal Regulations) 866.5910 as a multi-level panel plus a daily single-level control, to US laboratories running tumor-informed MRD assays.

This file carries no confidence tags. It cites them: [claim N: tag] reads claim N of the candidate file with the tag it carries there, and check_assessment.py refuses a citation that disagrees. It moves no score, no stage: and no status:.

How these gradings were made, disclosed rather than smoothed over. This is the first assessment of idea-003, stated on 2026-09-26 against pivot P2 clinical QC material at launch, the shape the loop record recorded the same day. Every consequence class below was set on 2026-09-26. That is after the Verifier tagged claims 1-178, including the one refuted claim this file cites claim 10refuted. For every pillar resting on those claims, the grading is reconstruction, not prospective. The limit on the damage is the same one the loop record uses: each class was set from what the binding chair said breaks, in its P2 clinical QC material at launch position file and in the register's "if it is false" column, and not from the tag. Only claims 179-192 were untagged when these classes were set. That covers the three this file appended (190-192) and eleven the business case and decks appended the same day.

Register coverage forced some of the rests on: lists wider than this file would have drawn them. The checker requires every one of the loop record's seventy-one live register rows to sit under some pillar. Each row is placed where its own "if it is false" column lands. Where that pulled in a claim the pillar would survive without, or an adverse verified finding that lifts a standing, the prose says so (B2 The native pooled-plasma matrix is a real, B6 A company can build the P2 fixed layer from… and B8 Something outside the lab's own preference… below). The alternative was to drop register rows, and that is the front end's call in the loop record, not this file's.

The bet

A laboratory running a tumor-informed MRD assay has to show, before it reports patients and then on every run, that the assay still detects tumor DNA at the fractions it claims. CLIA requires control procedures that monitor the complete analytic process claim 7verified. It names no material for doing so claim 76verified, claim 77verified. So labs today make their own contrived controls, or dilute the one incumbent product with sub-0.1% levels, which is purified DNA in buffer claim 111unconfirmed, into their own plasma. The bet is that such a lab pays $400-$900 a vial for a single-use daily control, and $1,500-$3,000 for a validation panel claim 149unconfirmed. The material is built in pooled donor plasma, so it goes through extraction alongside the patient samples claim 187unconfirmed. It carries a stated uncertainty at every level, and it is listed as a Class II (FDA device class II) QC device under product code SCU (subcutaneous), which FDA marks 510(k)-exempt claim 131verified. The recurring money is the daily control claim 151unconfirmed. That line is the reason P2 clinical QC material at launch chose a clinical-QC label over an RUO (research use only) one, and it is also the line the RUO label disclaimed. The buyers are the thirty to sixty US labs running or developing these assays claim 12unconfirmed. The structural attraction is that no incumbent sells a plasma-matrix level below 0.1% claim 188unconfirmed. The structural danger is that the whole bet sits on one sentence, and the sentence has to satisfy three parties at once. FDA has to read it as exempt. The physics has to support its per-vial claims. A tumor-informed lab has to be able to run a control whose variants are not its patient's.

The shapes it is not, each of which survives a different pillar failing:

• Development-phase-only RUO. Sell to kit makers, pharma/CRO groups and research labs only, and drop run-by-run QC. It survives B4 The P2 intended-use sentence (no exemption question) and most of B6 A company can build the P2 fixed layer from… (no QMS before revenue). It does not survive B5 It is a business. Enough labs, because it loses the recurring line, and its own weak point is the customer count claim 154unconfirmed.
• Buffer or defined-matrix format. This is the Seraseq MRD mix's shape. It survives B2 The native pooled-plasma matrix is a real and most of B7 Nothing stands between the product and a…'s plasma-rights leg, and Commercial expects to block it on differentiation.
• Floor at 0.01% VAF (variant allele frequency). It survives most of B3 Every level below 0.1% can carry a… at the bottom level, and Technology, Operations and Clinical all lean toward it.
• Measured 0.1% daily control. It survives B3 Every level below 0.1% can carry a…'s derived-value leg for the high-volume SKU, and competes head-on with Seraseq Complete AF0.1% claim 118unconfirmed.
• Daily-control-only launch. It survives the panel premium claim 153unconfirmed failing, as long as the control holds.
• License the method to an incumbent (Route D). It survives B5 It is a business. Enough labs and B6 A company can build the P2 fixed layer from…, because the incumbents already hold ISO (International Organization for Standardization) 13485 systems. It dies on B7 Nothing stands between the product and a…'s patentability leg claim 175unconfirmed.

Pillars

Eight. Each is a separate way this fails, graded against the business. B1 A laboratory running a tumor-informed MRD… and B5 It is a business. Enough labs are the two that nothing survives. Every other pillar has a named shape from ## Pivot options on the table waiting behind it, which is what grading it reshaping or costly means here.

• B1 A laboratory running a tumor-informed MRD… A laboratory running a tumor-informed MRD assay can and will run a purchased third-party control as its routine run-by-run control. That means designing and maintaining a dedicated panel against the control's certified variants, accepting that panel as meeting its control requirement, and paying for it instead of diluting incumbent material or making its own — consequence: fatal — graded: 2026-09-26 — rests on: claim 21unconfirmed, claim 162unconfirmed, claim 177unconfirmed, claim 185unconfirmed, claim 187unconfirmed, claim 190unconfirmed, claim 191unconfirmed — settled by: the REQ-0038 Run a short structured survey of 10-15 MRD…/REQ-0086 Extend the REQ-0038 survey instrument…/REQ-0098 Before the REQ-0038/REQ-0086 survey is…/REQ-0105 Before the REQ-0038/REQ-0086/REQ-0098… survey instrument fielded to 10-15 MRD lab directors, with one added question on how their run controls meet CLIA when every patient's panel differs; plus a literature read of published tumor-informed MRD validations for how their run-level controls are built (claim 190unconfirmed) — cost: cheap — in plain words: A cancer-testing laboratory has to be willing to buy an outside quality- control sample and run it with every batch, build and maintain a dedicated test panel for it, accept it as meeting its quality rules, and pay for it rather than make its own. If labs will not, there is no customer.
• B2 The native pooled-plasma matrix is a real The native pooled-plasma matrix is a real, paid-for difference. Spike fragments recover like endogenous cfDNA across the extraction kits labs use. The material is more commutable than buffer-format incumbents. Pooled plasma can be sourced consistently. Buyers pay a premium for it. No incumbent ships the same thing first — consequence: reshaping — graded: 2026-09-26 — rests on: claim 30unconfirmed, claim 31unconfirmed, claim 90verified, claim 92verified, claim 135unconfirmed, claim 153unconfirmed, claim 183unconfirmed, claim 188unconfirmed — settled by: a three-kit extraction-recovery comparison of spiked fragments against endogenous cfDNA in purchased pooled plasma (Technology's WP3), run before any commutability study, which is REQ-0109 Source and cost the commutability and…'s EP30 design — cost: project — in plain words: Making the control out of real pooled human plasma, rather than a buffer, has to be a difference labs can measure and will pay extra for, with no competitor shipping the same thing first. If not, the product competes on price with what already exists.
• B3 Every level below 0.1% can carry a… Every level below 0.1% can carry a defensible assigned value, and that value survives to the customer's bench. This includes the 0.03% daily control, assigned in one gravimetric step. It means a dilution-chain uncertainty within budget, enough molecules per vial summed over a stable clonal variant set, low enough background in the pool, independent confirmation by aggregate sequencing, stability, and three-lot consistency, at a level where an in-control run does not miss — consequence: reshaping — graded: 2026-09-26 — rests on: claim 32unconfirmed, claim 60unconfirmed, claim 63unconfirmed, claim 133unconfirmed, claim 134unconfirmed, claim 138unconfirmed, claim 139unconfirmed, claim 160unconfirmed, claim 166unconfirmed, claim 167unconfirmed, claim 168unconfirmed, claim 169unconfirmed, claim 170unconfirmed, claim 186unconfirmed — settled by: the pilot lot Technology specified in its REQ-0082 Specify the single pilot-lot bench… answer (WP0-WP2 and WP5), priced by REQ-0095 Price, as a cost and calendar with named… at a judgement band of USD 0.5-0.9M and 9-12 months to a pass/fail — cost: funded — in plain words: Every very dilute level of the control has to carry a value a lab can trust, and that value has to hold from the factory to the lab bench and from one production lot to the next. If it cannot, the product is not a reference material, and proving it is a funded, year-long piece of work.
• B4 The P2 intended-use sentence The P2 clinical QC material at launch intended-use sentence, the plasma matrix and the restricted label all sit inside 866.5910's 510(k) exemption. That means no neoplastic-disease withdrawal, no different-intended-use reading, "during assay validation" read as monitoring, and FDA accepting the special control's performance testing on LDTs, since no tumor-informed MRD assay is cleared — consequence: reshaping — graded: 2026-09-26 — rests on: claim 52unconfirmed, claim 53unconfirmed, claim 131verified, claim 132unconfirmed, claim 155unconfirmed, claim 156unconfirmed, claim 157unconfirmed, claim 161unconfirmed, claim 163unconfirmed, claim 189unconfirmed — settled by: the 513(g) request drafted as REQ-0081 Draft the 21 CFR 513, filed after Regulatory answers REQ-0104 P2's intended-use sentence names… (label restriction against the sentence) and REQ-0090 Under the 866.5910 special control (derived values, LDT sites) — cost: project — in plain words: The product has to fit inside an existing FDA exemption for laboratory quality-control materials, so that it can be sold without a clearance. If the FDA reads the intended use differently, the route becomes a clearance and the timeline and budget change.
• B5 It is a business. Enough labs It is a business. Enough labs, enough recurring control volume not held by one laboratory, and prices at or above the incumbent anchor clear the fully loaded cost, including a fixed quality-system layer, at planning volumes of about 50 panels and 2,000 vials a year — consequence: fatal — graded: 2026-09-26 — rests on: claim 12unconfirmed, claim 26unconfirmed, claim 47unconfirmed, claim 49unconfirmed, claim 125unconfirmed, claim 140unconfirmed, claim 149unconfirmed, claim 150unconfirmed, claim 151unconfirmed, claim 152unconfirmed, claim 154unconfirmed, claim 178unconfirmed, claim 192unconfirmed — settled by: the same survey as B1 A laboratory running a tumor-informed MRD… for price and volume, a written SeraCare quote (REQ-0087 Obtain a written quote from SeraCare/LGC…), a directory count of the labs (REQ-0037 Obtain a licensed lab-economics report on…), Operations' cost build-up at Commercial's volumes (REQ-0083 Produce the cost-of-goods build-up your…), and a count of Medicare MRD volume by laboratory from the cached CMS utilization file (claim 192unconfirmed) — cost: project — in plain words: There have to be enough laboratories, buying enough vials a year, at a price at or above today's incumbent, to cover the full cost including a permanent quality system. If the volumes or prices are too low, there is no business.
• B6 A company can build the P2 fixed layer from… A company can build the P2 clinical QC material at launch fixed layer from a standing start, or buy it from a contract manufacturer, on the budget and calendar claim 20unconfirmed implies. The layer is a part 820 QMS, the special-control study set and three production lots on independently sourced pools, plus the plasma volume and screening those lots consume — consequence: costly — graded: 2026-09-26 — rests on: claim 20unconfirmed, claim 33unconfirmed, claim 58verified, claim 59verified, claim 61verified, claim 141unconfirmed, claim 142unconfirmed, claim 158unconfirmed, claim 159verified, claim 171unconfirmed, claim 172unconfirmed, claim 173unconfirmed, claim 179unconfirmed, claim 180unconfirmed, claim 181unconfirmed — settled by: retrieval of FDA's 2007 cystic-fibrosis QC special-controls guidance (REQ-0102 Retrieve FDA's 'Class II Special Controls…), then Regulatory's answer on whether its studies must precede distribution (REQ-0096 Under 21 CFR 866.5910 as a 510), then one contract-manufacturer quote (REQ-0100 For each FDA-registered establishment that…) — cost: cheap — in plain words: A company can build, or contract out, the quality system, the required studies and the first three production lots on the budget and timeline assumed. If it costs much more, the same business is slower and more capital-hungry.
• B7 Nothing stands between the product and a… Nothing stands between the product and a lawful sale, and there is something to own. No incumbent patent reads on the material, the dilution chain or the labelled run-by-run use. The plasma, the spike line and the proxy DNAs can lawfully be sold, characterized and disclosed. The dilution-chain combination is patentable — consequence: reshaping — graded: 2026-09-26 — rests on: claim 143unconfirmed, claim 145unconfirmed, claim 146unconfirmed, claim 147unconfirmed, claim 148unconfirmed, claim 174unconfirmed, claim 175unconfirmed, claim 176unconfirmed — settled by: counsel's freedom-to-operate and patentability search over the P1 derived-value bottom levels/P2 clinical QC material at launch method (REQ-0027 Run a patent/prior-art landscape search, REQ-0097 Supplement REQ-0027's landscape to the…), with the three granted claim sets from USPTO full text first (REQ-0077 Attach from USPTO full text), and the supplier and line terms (REQ-0078 Name the pooled human donor plasma supplier, REQ-0085 Name the variant-bearing source material…, REQ-0108 Extend REQ-0085's licence enquiry for the…) — cost: project — in plain words: No competitor's patent covers the material, the dilution method or the labelled use; the plasma and the synthetic DNA can lawfully be bought and sold; and the method is patentable. If a patent blocks it, the product changes or needs a licence.
• B8 Something outside the lab's own preference… Something outside the lab's own preference pulls it toward characterized third-party material. Candidates are a payer technical assessment, the CAP checklist, the BLOODPAC consensus, or downstream test revenue large enough to fund QC spend — consequence: costly — graded: 2026-09-26 — rests on: claim 10refuted, claim 11unconfirmed, claim 15unconfirmed, claim 55unconfirmed, claim 57unconfirmed, claim 66verified, claim 68verified, claim 71verified, claim 76verified, claim 77verified, claim 136unconfirmed, claim 137unconfirmed, claim 144unconfirmed, claim 164unconfirmed, claim 165unconfirmed — settled by: four document retrievals already raised: the LCD L38779 text (REQ-0072 Retrieve active MolDX LCD L38779), the MolDX TA FAQs (REQ-0093 Retrieve Palmetto GBA's MolDX Technical…), the BLOODPAC full text (REQ-0103 Retrieve the full text), and the CAP MOL checklist (REQ-0028 Read ISO 17034) — cost: cheap — in plain words: Something beyond the lab's own preference, a payer rule, an accreditation checklist, an industry consensus, or test revenue big enough to fund it, has to push labs toward buying characterised outside controls. If nothing does, adoption is slower and the business thinner.

Standing

Rendered, never written: python3 engine/scripts/check_assessment.py --candidate idea-003 --render

Decision queue

Same render. Worst exposure first; inside a band, cheapest first.

Regrades

None. This is the first statement of this assessment. The earlier P0 the original shape and P1 derived-value bottom levels shapes were never assessed, so there is nothing to regrade.

What this does not say

• B8 Something outside the lab's own preference… is holed, and it does not mean what holed usually means. It rests on claim 10refuted, refuted at IR-20 claim 10 came back refuted. National Medicare-FFS (fee-for-service) volume for 0340U is 25,746 claim 68verified, not above 100,000. B8 Something outside the lab's own preference… is graded costly, so the matrix makes it high and not blocking. The file names IR-20 claim 10 came back refuted as the review that answered it, which chose commission and moved sizing to the bottom-up chain. An intelligent reader could argue B8 Something outside the lab's own preference… should be reshaping. If no external requirement pulls labs to characterized material, demand rests wholly on B1 A laboratory running a tumor-informed MRD…'s preference, and the P2 clinical QC material at launch sales pitch changes. It is graded costly because B1 A laboratory running a tumor-informed MRD… already carries that preference as fatal, and failure here changes what the product earns, not what gets built. Claims 11 and 57 sit under B8 Something outside the lab's own preference… only because they are live register rows. Commercial now treats them as cross-check legs, and B8 Something outside the lab's own preference… would survive without them. The four verified claims under B8 Something outside the lab's own preference… are all adverse or neutral to it. CLIA names no material for establishing sensitivity claim 76verified, and it asks for traceability only for calibrators, "if possible" claim 77verified. The LDT rule stays vacated claim 66verified, so no FDA pull exists either. claim 68verified is the corrected volume. They are here because register rows A20 TAM component 2, A22 FDA's 2024 laboratory developed test final… and A32 Assumption cite them, and because "nothing in CLIA pushes labs to third-party material" is B8 Something outside the lab's own preference…'s subject.
• B4 The P2 intended-use sentence at moderate is the reading here most likely to mislead. It is partial only because claim 131verified verified that SCU exists, is Class II and is exempt. Every claim about whether this sentence, matrix and label fall inside that code is unconfirmed. So is every other B4 The P2 intended-use sentence claim. Those include claim 52unconfirmed, claim 156unconfirmed and claim 157unconfirmed, which are the three findings Regulatory says would move it to blocks. The moderate band is the arithmetic. The census column is the correction.
• B6 A company can build the P2 fixed layer from… at low is the same effect, and stronger. Four of its fifteen claims are verified, and three of those verify that something is missing. claim 20unconfirmed names no volume claim 58verified. The file names no single-source dependency claim 59verified. Donor screening is unbudgeted claim 61verified. The fourth, claim 159verified, verifies that a QMS is required before first distribution. All four are adverse, and all four lift the standing. The instrument cannot read direction. B6 A company can build the P2 fixed layer from… still has its own grading question: costly holds only if a larger build is merely slower and dearer. The loop record lists "REQ-0083 Produce the cost-of-goods build-up your… shows the QMS layer breaks claims 20/50" as a trigger for the development-phase-only shape. That is a reshaping consequence, and a reader who thinks the fixed layer is large relative to a revenue of about $875K claim 178unconfirmed should argue B6 A company can build the P2 fixed layer from… up. If they do, it gets a row in ## Regrades, not a quiet edit.
• B2 The native pooled-plasma matrix is a real at moderate is lifted by two verified findings that point against it. B2 The native pooled-plasma matrix is a real is partial because claim 90verified and claim 92verified are verified. claim 90verified is a study's authors concluding that well-defined variants matter more than native-likeness for sensitivity and precision work. claim 92verified is a 15-site study in which some extraction kits preferentially recovered spiked cell-line DNA over endogenous plasma cfDNA, which is this product's construction. Both are here because register rows A8 Commutability against clinical patient… and A40 Across at least three cfDNA extraction… cite them, and both are evidence against the matrix being the difference. The matrix arithmetic cannot read direction, so the band went down from high on the arrival of bad news. Read B2 The native pooled-plasma matrix is a real as high. The table is left alone, because patching a rendering would turn it into a memory.
• B3 Every level below 0.1% can carry a… is graded reshaping and not fatal, and that is arguable. P1 derived-value bottom levels's mechanism rests on claim 134unconfirmed. If the derived levels cannot hold a value, P1 derived-value bottom levels fails for the reason P0 the original shape did. The gradings rest on the pivot options already on file: a 0.01% floor, and a measured 0.1% daily control. Both leave a product, but a less differentiated one. If claim 135unconfirmed fails as well, B2 The native pooled-plasma matrix is a real has also failed, and at that point the reshaped product is the buffer-format incumbent's shape. This file does not grade that combination. It is a question of how the pillars interact, and the standing arithmetic has no column for it.
• B1 A laboratory running a tumor-informed MRD… and B5 It is a business. Enough labs at critical are the ordinary condition of an unmade product. They are not a verdict. What the queue adds is that B1 A laboratory running a tumor-informed MRD…, the one about whether anyone can use the control at all, can be moved this month with a survey and a literature read. B5 It is a business. Enough labs needs vendor quotes on top.
• Claims 190-192 were surfaced here, and they are the gap this assessment found. No file asked whether a tumor-informed lab can meet its control requirement with a control run on a panel that is not the patient's. claim 185unconfirmed asks whether it will, not whether that satisfies CLIA/CAP (claims 190, 191). No file asked whether run-by-run volume is spread across labs or held by one (claim 192unconfirmed). claim 177unconfirmed counts labs, not runs, and one lab in ten by count could be almost no volume. None of these three is a finding. They are unverified propositions, appended so the Verifier can reach them. This file states no view on which way they come out.
• **Exposure is not truth, resolved is not a pass, and nothing here ranks idea-003 against idea-001 or idea-002.** evidence-gated is a state. It says the biggest open questions are knowable and what they cost. It does not say this opportunity is worth money. That is Tier 3.

Every claim in the candidate file, with the pillars that rest on it and the materiality it inherits from them. 90 of 192 are read by a pillar; the rest are context — true, possibly checked, and carrying nothing. That is not an insult and it is not an error: it is the only view in this repo that can tell you where the verification effort went.