Connector coverage

Before any interpretation, every analysis runs a full data-connector sweep. The sweep is two tiers: a company tier called once per ticker and reused by every program under it, and a program tier called per program. A program with any mandatory connector left not called is provisional and cannot be signed off — this page is the audit trail. Each ticker's price cache appears below as a derived row, read straight off the committed cache file on disk, not as a state any connector call recorded.

As of 2026-08-20

TLSA Cleared

Company tier

Swept 2026-08-10 · called once, reused by every program

BPIQ:fetch_company_info	Called
BPIQ:fetch_company_drugs	Called
BPIQ:fetch_company_historical_catalysts	Called
BPIQ:fetch_company_options_data	Called
BPIQ:fetch_company_insider_transactions	Called
BPIQ:fetch_company_press_releases	Called
BPIQ:fetch_company_hedge_fund_holdings	Called
EDGAR:submissions	Called
EDGAR:xbrl_companyconcept_dei_EntityCommonStockSharesOutstanding	Called
FINRA:consolidatedShortInterest	Called
Yahoo:options_chain	Called

Program tier — foralumab-na-spms

BPIQ drug id 13497 · CLEARED

CT.gov:search_trials+get_trial_details	Called
CT.gov:search_investigators	Called
PubMed:search_articles+get_article_metadata	Called
Open Targets:search_entities Rate limit exceeded for client: global	Blocked
ChEMBL:compound_search	Called
web_search:x4	Called
EuropePMC:search	Called
EDGAR:full_text_search	Called
CTIS:search	Called
NIH RePORTER:projects_search	Called

CD3 genetics unverified - Open Targets BLOCKED on three attempts, FOURTH consecutive sweep

PubMed still omits the peer-reviewed na-SPMS paper (DOI 10.1212/NXI.0000000000200543) from its 15 foralumab hits, fourth sweep running; Europe PMC returns it as PMC12815464 with the full 26-author list. The new optional row closed a gap three sweeps could only flag. Full text still not read.

Europe PMC returns 88 hits where PubMed returns 15 on the identical query - the two indexes are not interchangeable and this program now uses both

CT.gov still discloses NO investigator for the pivotal trial: no overall officials, contacts null on all seven sites, search_investigators finds nobody on NCT06292923. Every kol investigator comes from a sibling trial or the publication record.

CT.gov NCT06292923 still contradicts itself on PET coverage ('at the PET sub-study site' vs 'Subjects at all sites'); evaluable n unknown. Fourth sweep.

Enrollment conflict persists: CT.gov n=54 vs company 48 / 'up to 48'

Primary-endpoint description conflict: CT.gov lists three co-primaries, the company names only PET

Catalyst date disagrees across four sources; readout.disagreement UNRESOLVED

NIH RePORTER queried on Weiner and Chitnis but NOT on Singhal - his federal-funding status is unchecked rather than clear

SPMS prevalence source reads 22.42 per 100,000 pooled against the 27-45 range B.3a is built on; correcting would cut peak sales ~20%. The non-active share of SPMS remains unverified on a fourth sweep.

data/benchmarks/readout-lag.json still holds no observation, so the modelled readout uses rule 34's stated default lag

The committed price cache ends 2026-08-05 ($0.99), three trading days behind the $1.00 locked spot; every price-derived figure comes from the cache, the spot does not. Refreshing the cache was outside this lane.

THREE NEW v5.6.0 MANDATORY ROWS CALLED FOR THE FIRST TIME. EDGAR submissions (CIK 0001723069, 480 recent filings): nine 2026 Section 16 filings, five Form 4 purchase transactions Aug 2026, F-3 333-286064 shelf effective 2025-03-27 with $250M base / $100M Jefferies ATM. EDGAR XBRL: 120,507,401 shares filed at 2025-12-31 (FY2025 20-F) - annual-only series as expected for an FPI. FINRA settlement 2026-07-15: 1,245,738 shares short, days-to-cover 4.62, -20.43% vs prior; settlement 2026-07-31 returned zero bytes (publication lag, not a missing ticker).
fetch_company_insider_transactions: no rows again - TENTH empty attempt across FOUR sweeps - on a day EDGAR shows five Form 4 purchase transactions filed 2026-08-05/06/07. The EDGAR submissions row now carries the flow as structured, dated, priced, coded data. The connector gap is unchanged; what changed is that it no longer costs the analysis anything.
fetch_company_info: the previous-close market_cap bug DID NOT REPRODUCE this sweep (127,259,128 / 1.00 last price = the implied 127.26M share count exactly; / 0.97 would give 131.2M). max_52_week_position (0.07831) likewise matches the last-price computation. The bug is INTERMITTENT, which is worse than stable.
fetch_company_info: ttm_burn returned -8,749,000 where every prior sweep returned +8,749,000 (sign flip, unchanged magnitude); enterprise_value crossed from $2.2M ABOVE market cap (2026-08-07) to $2.3M BELOW it (2026-08-10) with no financing event between. monthly_burn, qtr_burn, finance_updated_at null again.
Yahoo options cross-check AGREES the chain is unreadable: same three strikes ($2.50/$5.00/$7.50), same four expiries, no strike within 2.5x of spot. Its own IV artifacts differ from BPIQ's - calls pinned at 0.500005, puts at 0.00001, against BPIQ's 0.01488 floor. Two vendors, two sentinel families, one unreadable chain. Yahoo shows zero front-expiry OI where BPIQ shows 29-1,626 contracts.
Company sweep of 2026-08-07 confirmed still CURRENT per Company-sweep currency: press feed re-read to date_updated 2026-08-10 (most recent item 2026-08-08), no earnings release and no financing intervening. Historical catalysts, options and hedge-fund rows reused on that basis.

Price cache

Derived from data/prices/TLSA.json — not a sweep row, so it carries no CALLED/BLOCKED state

First bar	2021-08-06
Last bar	2026-08-19
Bars	1264
Fetched	2026-08-19

ABOS Cleared

Company tier

Swept 2026-08-07 · called once, reused by every program

BPIQ:fetch_company_info	Called
BPIQ:fetch_company_drugs	Called
BPIQ:fetch_company_historical_catalysts	Called
BPIQ:fetch_company_options_data	Called
BPIQ:fetch_company_insider_transactions	Called
BPIQ:fetch_company_press_releases	Called
BPIQ:fetch_company_hedge_fund_holdings	Called
BPIQ:fetch_company_earnings_transcript	Called

Program tier — sabirnetug-alzheimers

BPIQ drug id 14145 · CLEARED

CT.gov:search_trials+get_trial_details	Called
PubMed:search_articles+get_article_metadata	Called
Open Targets:search_entities	Called
ChEMBL:compound_search	Called
web_search:x4	Called

Open Targets search_entities succeeded on the FIRST attempt this sweep (both prior sweeps needed the retry); the follow-up graphql association query stayed BLOCKED across three attempts, third consecutive sweep - quantitative APP-AD association strength remains UNVERIFIED.

CT.gov NCT06335173 discloses no overall officials and no site investigators (all 68 location contacts null); the kol investigator panel is built from the program's peer-reviewed publications instead.

PubMed get_full_text_article returned PMC13130801 (Trials qualitative paper) WITHOUT its Declarations/COI section; PMC12184067's COI section did come through (it reads 'None'). Inconsistent back-matter retrieval is recorded per-row in kol.conflicts_checked.

Halozyme ENHANZE royalty (single-digit on net sales of the SC form) confirmed this sweep from the 2023-11-05 agreement announcement, RESOLVING the gap the 2026-08-04 sweep declared; the JCR royalty rate remains unconfirmed (EBD programme only, not sabirnetug).

No company statement names CTAD 2026 for the topline (press feed + targeted web search checked), so the congress readout source is an explicit null with a note.

Full company-tier re-run (previous stamp 3.7.0 was below stale_before 5.0.0). fetch_company_drugs: same 4 rows as 2026-08-04; only 14145 has_catalyst true; note field now carries six consecutive 'Late 2026' reiterations, zero slips.
fetch_company_info: price $2.28; financial fields still stamped 2026-03-31 (no newer balance sheet; Q2 2026 results scheduled 2026-08-12, announced 2026-08-05). Market-cap share count CORRECTED upstream since 2026-08-04 (now matches the 10-Q's 72,227,580 to within six shares) but still computed on the previous close; max_52_week_position reproduces the previous-close bug at 0.43568.
fetch_company_press_releases: 207 items to 2026-08-05, read from the saved overflow file. One material new item: Q2 2026 results date 2026-08-12.
fetch_company_insider_transactions: 233 rows to 2026-06-17, no new row since 2026-08-04; full-history behaviour reproduces.
fetch_company_options_data: still three strikes ($2.50/$5.00/$7.50), no ATM option on a $2.28 stock, December OI 153, straddle not computable; IV no longer byte-identical across expiries but floor artifacts persist. New: 340 contracts of Jan-2027 call volume, first visible positioning across the readout window.
fetch_company_hedge_fund_holdings: 2026Q2 unchanged (RA 20.99M / ADAR1 2.16M / Ikarian 1.05M, all implying $2.36). fetch_company_earnings_transcript: still the Q1 2026 call (2026-05-12); Q2 call is 2026-08-12.

Price cache

Derived from data/prices/ABOS.json — not a sweep row, so it carries no CALLED/BLOCKED state

First bar	2021-08-06
Last bar	2026-08-19
Bars	1264
Fetched	2026-08-19

ZNTL Cleared

Company tier

Swept 2026-08-12 · called once, reused by every program

BPIQ:fetch_company_info	Called
BPIQ:fetch_company_drugs	Called
BPIQ:fetch_company_historical_catalysts	Called
BPIQ:fetch_company_options_data	Called
BPIQ:fetch_company_insider_transactions	Called
BPIQ:fetch_company_press_releases	Called
BPIQ:fetch_company_hedge_fund_holdings	Called
EDGAR:submissions	Called
EDGAR:xbrl_companyconcept_shares_outstanding	Called
FINRA:consolidatedShortInterest	Called

Program tier — azenosertib-cyclin-e1-proc

BPIQ drug id 17802 · CLEARED

CT.gov:search_trials+get_trial_details	Called
PubMed:search_articles+get_article_metadata	Called
Open Targets:search_entities Rate limit exceeded for client: global	Blocked
ChEMBL:compound_search	Called
web_search:x4	Called

CCNE1/WEE1 human genetics unverified (Open Targets blocked, 4 attempts — the documented standing block); target validation rests on published papers

CT.gov NCT05128825 primary_completion_date (2026-12-01) predates the 2026-08-06 slip and points 5-7 months earlier than company guidance — recorded as an UNRESOLVED readout disagreement; watch the registry's next edit

No peer-reviewed DENALI clinical data exists; every clinical figure is conference- or press-release-sourced (PubMed 17 hits, all metadata retrieved, 16 relevant)

PubMed metadata carries no COI field — kol conflicts_checked entries record searches made, never certified independence; the DENALI registry names no overall officials and one site contact, so the investigator panel is built from the sponsor biomarker paper's clinical co-authors

ChEMBL field errors persist (oral:false, max_phase 2); neither used

Recurium royalty wording conflicts across filing descriptions ('mid- to high-single digit' this sweep vs 'low to mid' earlier); sublicensing percentage not established

FDA Fast Track grant date conflict persists (2026-04-21 BPIQ vs January 2025 earlier source)

Price cache ends 2026-08-05 (pre-slip); spot and the -16% reaction read from the 2026-08-12 BPIQ quote, not committed bars

Regulatory tier not called — catalyst is a trial readout, not a regulatory event

fetch_company_info: finance_updated_at 2026-06-30 — the Q2 results staleness flagged on 2026-08-04 has resolved; market_cap and max_52_week_position both reproduce the previous-close bug exactly ($294.6M / $4.11 = 71.67M shares, matching the EDGAR-filed 71,671,914 to 0.01%)
fetch_company_drugs: twelve rows, identical id set to 2026-08-04; row 17802's catalyst_date_text moved from 'YE 2026' to 'H1 2027' (placeholder 2027-06-30) with note '08/06/26: DENALI FDA aligned; ESMO OS update upcoming; Part 2 topline now H1 2027'; is_big_mover and is_high_mgmt_interest now true
fetch_company_historical_catalysts: eight rows, unchanged; the 2026-08-06 slip reaction is an earnings item and not in this feed — read from the price cache's last bar (4.95, 2026-08-05) against the 2026-08-12 quote (~-16%)
fetch_company_insider_transactions: 250 rows read in full from the saved file; corrects the 2026-08-04 sweep's dating of the last open-market insider buy (Bruns $2.28 is 2025-02-06, not 2026-02-06; the Feb-2026 rows are tax disposals)
fetch_company_hedge_fund_holdings: the v5.6.1 filing-quarter labelling bug is CONFIRMED on ZNTL — '2026Q2' value/shares = $2.34 = 2026-03-31 close, '2026Q1' = $1.35 = 2025-12-31, '2025Q4' = $1.51 = 2025-09-30 (matched against data/prices/ZNTL.json). No 13F visibility past 2026-03-31; the previous sweep's 'flat in 2026Q2' claim was the label bug
EDGAR submissions: 538 filings; 10-Q + 8-K 2026-08-06, new Schedule 13G 2026-07-30 (holder unread), live S-3 shelf 2025-03-26, no 424B5 since 2023-06-15
EDGAR XBRL dei:EntityCommonStockSharesOutstanding: 71,671,914 at 2026-08-03 (10-Q cover)
FINRA consolidatedShortInterest: 2026-07-31 settlement 5,578,646 shares (7.78% of filed count), days-to-cover 4.16; built 23% mid-June to mid-July then covered 9% before the slip. BPIQ short_float (11.76%) superseded
options: last listed expiry 2027-01-15 covers only the first two weeks of the H1 2027 readout window; near-spot quotes bracket 16-178%; the deep-ITM $2.50 Jan-27 call OI block shrank 4,302 -> 1,230

Price cache

Derived from data/prices/ZNTL.json — not a sweep row, so it carries no CALLED/BLOCKED state

First bar	2021-08-06
Last bar	2026-08-19
Bars	1264
Fetched	2026-08-19

IMUX Cleared

Company tier

Swept 2026-08-11 · called once, reused by every program

BPIQ:fetch_company_info	Called
BPIQ:fetch_company_drugs	Called
BPIQ:fetch_company_historical_catalysts	Called
BPIQ:fetch_company_options_data	Called
BPIQ:fetch_company_insider_transactions	Called
BPIQ:fetch_company_press_releases	Called
BPIQ:fetch_company_hedge_fund_holdings	Called
BPIQ:fetch_company_earnings_transcript	Called
EDGAR:submissions	Called
EDGAR:companyconcept	Called
FINRA:consolidatedShortInterest	Called
Yahoo:options_chain	Not called

Program tier — vidofludimus-relapsing-ms

BPIQ drug id 16219 · CLEARED

CT.gov:search_trials+get_trial_details	Called
PubMed:search_articles+get_article_metadata	Called
Open Targets:search_entities Rate limit exceeded for client: global	Blocked
ChEMBL:compound_search Tool 'compound_search' exceeded 30.0s server timeout (attempts 1-3); Internal error: Error calling tool 'compound_search': HTTP error in tool compound_search: 500 from upstream API (attempts 4-6)	Blocked
web_search:x4	Called
EDGAR:full_text_search	Called
EuropePMC:search	Not called
CTIS:search	Not called

Open Targets search_entities BLOCKED on four attempts across the full retry policy, reproducing the standing block in framework/02-connectors.md. The claim left unverified is independent genetic target validation for DHODH and NR4A2; target validation rests on teriflunomide's approval instead, so nothing depends on the call.

ChEMBL compound_search BLOCKED on six attempts across the full retry policy - a regression, since the 2026-08-04 sweep obtained a live result. The molecule-identity, physicochemical and selectivity figures are carried from that call with provenance stated on the chembl block rather than re-tagged as verified today.

BPIQ intra_day_price_change_percent does not reconcile with the committed price cache on any of the seven rows the cache covers and reverses the sign of 2024-10-22. This is the largest finding of the sweep and it changed two elements of the analysis: the nearest-comparable-reaction reading and one positive-scenario anchor.

Readout sources disagree and are left UNRESOLVED: company guidance says topline by 2026-12-31; the registry's 2026-11 primary completion plus rule 34's default two-to-four-month lag gives 2027-01-30 to 2027-03-30.

ENSURE's primary endpoint is time to first relapse (hazard ratio) while every comparator the market anchors on reports annualised relapse rate. Not interchangeable, no defensible conversion without patient-level data. Flagged, not resolved.

CT.gov lists no overall officials and no site contacts for either ENSURE trial, so search_investigators returns nothing; the KOL investigator panel is built from the pivotal Phase 2 publication cross-referenced against the ENSURE site lists.

No conflict-of-interest statement is available through the PubMed connector; every conflicts_checked entry records an author-list and affiliation check only.

PubMed returned 47 hits, above the 15-article threshold. Metadata retrieved for the 20 most recent plus four targeted MS reviews; 27 records not individually inspected.

EMPhASIS enrolment conflict: CT.gov n=210 vs company and publication n=268. Both carried, neither chosen.

Trial-geography conflict: the company says 15 countries, the two registry records list 24. Not resolved.

CALLIPER's registry record (NCT05054140) still reads status UNKNOWN on re-check, so its result is taken from company reporting.

Fenebrutinib's regulatory filing status could not be sourced - no filing acceptance, priority review or action date found - and the date its 2023-11-30 partial clinical hold was lifted could not be sourced at all.

The tolebrutinib Complete Response Letter is dated December 2025 by one third-party source and April 2026 by another; neither is primary and the date is not resolved.

The 60% US treatment rate underlying the peak-sales build is unverified and is the weakest input; the single $30,000 net price is a modelling judgement anchored on a verified 2021 class figure.

The ENSURE powering assumptions have never been disclosed numerically, which is what sets the width of the probability band.

No regulatory designation could be confirmed either way; the absence is the result of a search, not a confirmed negative from a regulator.

A fifth web search beyond the mandated four was run, to resolve fenebrutinib's clinical-hold history.

fetch_company_drugs: 8 rows, unchanged in number and in has_catalyst since 2026-08-04. Exactly one row (16219) carries a catalyst, so lib/clustering.mjs returns CLEAN with an empty conflict list.
fetch_company_historical_catalysts: 8 rows, 2021-02-18 to 2026-02-04. The call succeeded but intra_day_price_change_percent does not reconcile with data/prices/IMUX.json on any of the 7 rows the cache covers, and on 2024-10-22 it reverses the sign of the day. C.7 recomputed from the cache throughout. open_price_gap_percent, by contrast, reproduces exactly on all 7 rows.
fetch_company_insider_transactions: 250 rows (2017-09-13 to 2026-07-05). Exceeded the tool output limit and was read in full from the saved file. No new transaction since the previous sweep.
fetch_company_press_releases: 283 items (2021-09-22 to 2026-08-02), date_updated 2026-08-10. Exceeded the tool output limit and was read in full from the saved file.
fetch_company_options_data: chain dated 2026-08-10, 4 expiries (2026-08-21, 2026-09-18, 2026-10-16, 2027-01-15). No expiry falls inside the readout window. The IV floor artifact (0.01488) is present on six contracts.
fetch_company_hedge_fund_holdings: 18 filing-quarter blocks, 2022Q2 to 2026Q2. The v5.6.1 filing-quarter labelling bug is confirmed exactly: value/shares on the '2026Q2' block gives $11.10 post-split, the 2026-03-31 close to the cent; '2026Q1' gives $5.34, the 2025-12-31 close; '2025Q4' gives $8.82, the 2025-09-30 close. Every quarter attribution shifted back one block.
fetch_company_earnings_transcript: returned an empty result, which counts as CALLED.
EDGAR submissions: CIK 0001280776, 816 recent filings. No Q2 2026 Form 10-Q as of 2026-08-11 (deadline ~2026-08-14). Three new Schedule 13G/13G/A filings since the last sweep.
EDGAR XBRL companyconcept: dei:EntityCommonStockSharesOutstanding = 13,621,526 at cover date 2026-04-30, replacing the figure the previous sweep derived from BPIQ's buggy market_cap. us-gaap Cash, NetIncomeLoss, OperatingExpenses, R&D and G&A also read.
FINRA consolidatedShortInterest: settlement 2026-07-15, 1,413,871 shares short, ADV 149,504, days-to-cover 9.46. The 2026-07-31 settlement returned empty, which is publication lag, not a missing ticker.
Yahoo options chain: optional row, deliberately skipped. The BPIQ chain is already unusable for a reason no per-strike cross-check would change - no listed expiry falls inside the readout window at all.
Price-cache caveat (derived, not a sweep row): data/prices/IMUX.json was last fetched 2026-08-06 and ends 2026-08-05, so it does not cover the three sessions before this sweep.

Price cache

Derived from data/prices/IMUX.json — not a sweep row, so it carries no CALLED/BLOCKED state

First bar	2021-08-06
Last bar	2026-08-19
Bars	1264
Fetched	2026-08-19

IVA Cleared

Company tier

Swept 2026-08-12 · called once, reused by every program

BPIQ:fetch_company_info	Called
BPIQ:fetch_company_drugs	Called
BPIQ:fetch_company_historical_catalysts	Called
BPIQ:fetch_company_options_data	Called
BPIQ:fetch_company_insider_transactions	Called
BPIQ:fetch_company_press_releases	Called
BPIQ:fetch_company_hedge_fund_holdings	Called
EDGAR:submissions	Called
EDGAR:xbrl_companyconcept_shares_outstanding	Called
FINRA:consolidatedShortInterest	Called

Program tier — lanifibranor-mash

BPIQ drug id 16301 · CLEARED

CT.gov:search_trials+get_trial_details	Called
PubMed:search_articles+get_article_metadata	Called
Open Targets:search_entities Rate limit exceeded for client: global	Blocked
ChEMBL:compound_search	Called
web_search:x4	Called

PPARA/PPARG/PPARD human genetics unverified (Open Targets blocked, 3 attempts, both sweeps)

Patent estate now disclosed via the 20-F (six US patents, expiries by 2026-12, 2027-09, 2035-06, 2039-11, 2041-12 ex-PTE; ~150 patents in ~55 jurisdictions) but read only via a web-search excerpt; expiry-to-claim-type mapping unverified

PubMed: 74 hits; metadata this sweep for 16 (12 most recent + 4 targeted); nothing new bears on the readout; older records not individually inspected

CT.gov returns a null secondary_outcomes array for NCT04849728 (re-confirmed); key secondaries from the detailed description

Registry lists no overall officials and null site contacts for NCT04849728 - the KOL investigator panel is built from the pivotal publications instead, and PubMed metadata carries no COI field, so conflicts_checked records searches made, not certified independence

Enrolment conflict persists: CT.gov 1,000 vs company 1,009; ~410-patient exploratory cohort absent from the registry figure

BPIQ internal inconsistency persists: catalyst_date_text 'H2 2026' vs note 'Q4 2026' (2026-05-26)

CT.gov primary completion 2027-09-30 tracks the Part B safety outcome, not the Part A topline - decomposed, not a conflict

Published literature swaps NATIVE/NATiV3 NCT numbers in both directions (two meta-analyses and the company's own biomarker paper)

RESOLVE-IT elafibranor failure still stated from general knowledge; no primary source retrieved

The price cache's 2026-08-11 bar carries a null close (Yahoo artifact); the session's $4.72 close confirmed against the matching BPIQ quote

fetch_company_info: finance_updated_at null; cash 230,870,000 is still a EUR figure in a USD field, dated 2025-12-31; ttm_burn's sign flipped negative since 2026-08-04, magnitude unchanged. NEW: the whole quote block matches the price cache's previous-session bar (2026-08-11), so BPIQ's 'live' quote lags one session on this ticker - the 2026-08-04 sweep's $4.26 spot was actually the 2026-08-03 close
fetch_company_info: the volume field is unreliable - 19,706 reported for 2026-08-04 vs 1.02M in the price cache; the previous 'thin ADS line' conclusion was an artifact
fetch_company_drugs: four rows, identical id set to 2026-08-04
fetch_company_historical_catalysts: six rows, byte-identical to 2026-08-04; still nothing comparable to a Phase 3 efficacy readout
fetch_company_insider_transactions: empty twice; EDGAR submissions confirm no Form 3/4/5 exist (French FPI), so the empty is structural
fetch_company_press_releases: 218 items to 2026-08-06. Read from the saved file: everything from 2026-06-25 in full plus a keyword scan of 2026-01-01 to 2026-06-24 titles; the pre-2026 feed was read in full on 2026-08-04 and is unchanged
fetch_company_hedge_fund_holdings: the v5.6.1 filing-quarter labelling bug is CONFIRMED on IVA - the '2026Q2' block's value/shares ratios equal the 2026-03-31 close ($5.55) and '2026Q1' the 2025-12-31 close ($4.65). Labels shifted back one quarter; no 13F visibility past 2026-03-31 exists yet
EDGAR XBRL dei:EntityCommonStockSharesOutstanding: 193,236,978 at 2025-12-31 (20-F) - annual-only for an FPI, superseded by the company's monthly statement of 236,280,202 at 2026-07-31 (via press feed)
FINRA consolidatedShortInterest: 2026-07-31 settlement 2,195,099 ADSs, days-to-cover 2.66, up 54% in six weeks - a pre-event short build; 0.93% of shares outstanding on the filed count

Price cache

Derived from data/prices/IVA.json — not a sweep row, so it carries no CALLED/BLOCKED state

First bar	2021-08-06
Last bar	2026-08-19
Bars	1264
Fetched	2026-08-19

OLMA Cleared

Company tier

Swept 2026-08-12 · called once, reused by every program

BPIQ:fetch_company_info	Called
BPIQ:fetch_company_drugs	Called
BPIQ:fetch_company_historical_catalysts	Called
BPIQ:fetch_company_options_data	Called
BPIQ:fetch_company_insider_transactions	Called
BPIQ:fetch_company_press_releases	Called
BPIQ:fetch_company_hedge_fund_holdings	Called
BPIQ:fetch_company_earnings_transcript	NOT CALLED (OPTIONAL)
EDGAR:submissions	Called
EDGAR:xbrl_companyconcept_shares_outstanding	Called
FINRA:consolidatedShortInterest	Called
Yahoo:options_chain	NOT CALLED (OPTIONAL)

Program tier — palazestrant-er-positive-mbc

BPIQ drug id 15137 · CLEARED

CT.gov:search_trials	Called
CT.gov:get_trial_details	Called
PubMed:search_articles	Called
PubMed:get_article_metadata	Called
OpenTargets:search_entities	Called
ChEMBL:compound_search	Called
web_search:peak_sales	Called
web_search:competitive	Called
web_search:exclusivity_royalty	Called
web_search:analyst	Called

CT.gov NCT06016738 is demonstrably unmaintained (status RECRUITING, primary completion 2026-06-30 passed, both contradicting the sponsor's 2026-08-10 enrollment-completion announcement); the registry offers no independent check on the Q1 2027 guidance.

Database lock never disclosed; enrollment completion announced without a date - Q1 2027 guidance cannot be checked arithmetically from outside.

No patent number, expiry, licence or royalty for palazestrant found (the Aurigene royalty attaches to OP-3136); market exclusivity unassessable.

Elacestrant price anchor internally contradictory (>10x spread between sources, neither net); US patient count under peak sales unverified.

No named independent KOL voice on the wild-type endpoint located; kol.judgement recorded as UNKNOWN rather than guessed.

Company-tier refresh 2026-08-12: prior company.md (v3.5.0, swept 2026-08-04) was below stale_before, so the tier was re-run in full rather than reused.
BPIQ market_cap combined both documented bug families at once: stale share count (87.35M) x previous close (11.94). Recomputed from EDGAR XBRL 87,497,607 shares.
BPIQ note field on drug 15137 was REPLACED, not appended - the 15-month guidance history is gone from the feed; only the 2026-08-10 entry survives.
Hedge-fund 13F label shift confirmed on OLMA: '2026Q2' block value/shares ratios equal the 2026-03-31 close ($14.91).
FINRA 2026-07-31 settlement was already published at this sweep (unlike the corpus's 2026-08-10 test): 15,963,215 shares short, 18.21 days to cover, fifth consecutive rise.
data/prices/OLMA.json bar 2026-08-11 carries a null close with real volume (2,670,582) - the -12.6% slip-reaction session; cache-derived figures resolve to the pre-slip 11.94 close until the bar self-heals on the next fetch.
globenewswire.com timed out on three WebFetch attempts for the Q2 2026 release; read in full via BioSpace mirror, cross-checked against the Q2 10-Q, which agrees.

Price cache

Derived from data/prices/OLMA.json — not a sweep row, so it carries no CALLED/BLOCKED state

First bar	2021-08-06
Last bar	2026-08-19
Bars	1264
Fetched	2026-08-19

GNLX Cleared

Company tier

Swept 2026-08-11 · called once, reused by every program

BPIQ:fetch_company_info	Called
BPIQ:fetch_company_drugs	Called
BPIQ:fetch_company_historical_catalysts	Called
BPIQ:fetch_company_options_data	Called
BPIQ:fetch_company_insider_transactions	Called
BPIQ:fetch_company_press_releases	Called
BPIQ:fetch_company_hedge_fund_holdings	Called
BPIQ:fetch_company_earnings_transcript	Called
EDGAR:submissions	Called
EDGAR:companyconcept	Called
FINRA:consolidatedShortInterest	Called

Program tier — olvi-vec-platinum-resistant-ovarian

BPIQ drug id 17406 · CLEARED

CT.gov:search_trials+get_trial_details	Called
PubMed:search_articles+get_article_metadata	Called
Open Targets:search_entities	Called
ChEMBL:compound_search	Called
web_search:x4	Called

STATE CHANGE SINCE 2026-08-05: Open Targets moved from BLOCKED to CALLED. The previous sweep recorded four consecutive 'Rate limit exceeded for client: global' errors and asserted nothing either way. This sweep it answered: 'olvimulogene nanivacirepvec' resolves to exactly one entity, CHEMBL4594308 typed 'drug', with NO target entity; 'vaccinia oncolytic virus' and 'platinum-resistant ovarian cancer' each returned a legitimate empty; and a control query for 'ovarian cancer' returned three disease entities, confirming the connector was live rather than degraded. The finding is now positive rather than absent: there is no genetic target-validation evidence to have, because the agent has no molecular target, and that has been checked. A.5 still scores target validation Low, but for a stated reason. One follow-up probe later in the sweep hit the global throttle again, so the connector remains intermittent.

ChEMBL compound_search recovered ONLY under the full retry policy - three 30s timeouts and one HTTP 500 before the fourth attempt returned. Recorded CALLED because it ultimately returned. A shorter retry policy would have recorded BLOCKED and looked like a different finding; this is a concrete case where the policy in 02-connectors.md earned its keep. A fifth call on the alias 'Olvi-Vec' returned a legitimate empty (count 0), so that name is simply not indexed.

ChEMBL still returns no usable characterisation for a live virus: CHEMBL4594308 carries molecule_type 'Gene', structure_type NONE, max_phase 3, orphan 0 and synonyms GL-ONC1 / GLV-1h68, but no molecular properties, no SMILES, no InChI and no bioactivity records. No selectivity or off-target claim is made anywhere in the document.

PubMed hit count is NOT comparable to the previous version's. That analysis recorded 9 hits; the broad query used here (Olvi-Vec OR olvimulogene OR GL-ONC1 OR GLV-1h68) returns 57, because it includes the laboratory name and its long preclinical literature. Metadata was pulled on the clinically relevant subset rather than on all 57 - a departure from the 'fewer than 15, pull all' rule, disclosed rather than glossed. A second narrower query (oncolytic virus / vaccinia x ovarian x platinum resistance, 2025 onward, 24 hits) supplied the KOL independent voices.

LAST PATIENT IN AND DATABASE LOCK ARE STILL UNDISCLOSED - absent from the ClinicalTrials.gov record, from every PubMed record read and from all 137 press-feed items. This is what makes readout.disagreement UNRESOLVED rather than CONSISTENT.

The registered status and the guidance are now in OPEN tension, not merely 'in spirit'. NCT05281471 read RECRUITING on 2026-08-11 with all 32 sites individually listed as recruiting, six weeks into the half-year the company has guided topline for. The 2026-08-05 record described the three date sources as agreeing; this sweep records disagreement UNRESOLVED, because a topline inside H2 2026 would have to be announced essentially simultaneously with a primary completion dated 2026-12, leaving no time for database lock and analysis. Neither reading is picked and neither is averaged.

The pivotal design paper's correction is STILL not retrievable: Int J Gynecol Cancer 2023;33(10):1672, DOI 10.1136/ijgc-2023-004812corr1, indexed with the title 'Correction:' and no abstract, confirmed again this sweep. What was corrected is unknown. Every design figure used (2:1, n=186, 127 events, PFS primary) is from the original abstract and is corroborated against the ClinicalTrials.gov record.

The ovarian efficacy literature is still not independent, and the newest paper does not change that. The 2026 translational analysis (Yu et al., Gynecol Oncol Rep 2026;66:102145, DOI 10.1016/j.gore.2026.102145, PMID 42383154) is new in published form since the last sweep and is the most direct published support the unproven step has - but its first author is a Genelux employee and its second is the trial's national principal investigator. Robert Holloway authors all four ovarian papers and is the Phase 3 national PI. The one genuinely independent mechanistic study remains the MSK intrapleural Phase 1.

NEW CONTEMPORANEOUS NEGATIVE: the IPROC platform trial (MacKay et al., Cancer Treat Res Commun 2026;48:101260, DOI 10.1016/j.ctarc.2026.101260) reported durvalumab plus two different ADCs in platinum-resistant high-grade serous ovarian cancer with ZERO responses across both sub-studies, neither proceeding past stage 1, and no differences in inferred immune-cell frequencies between pre- and on-treatment samples. A different mechanism from Olvi-Vec, so weak evidence about this agent specifically, but direct evidence that immune activation in this population does not reliably convert into benefit. One of the two inputs behind lowering the probability from 40% to 35%.

KOL panel assembled for the first time on this program (kol block, new since v5.5.0). Six investigators recorded from the registry and the programme's own publications; three independent voices recorded from the 2025-2026 literature. Conflict checks are AUTHORSHIP-LEVEL ONLY: journal disclosure appendices were not opened for any named investigator, and 26 of the 32 named site contacts were not conflict-checked at all. Two competitor relationships were found and recorded - Holloway and Thaker both co-author the Verastem RAMP 201 publication, and Holloway also co-authors the GSK RUBY publication.

One prominent voice was found and DELIBERATELY EXCLUDED from the independent panel: the Hemminki editorial on oncolytic adenoviruses in platinum-resistant ovarian cancer (DOI 10.1080/14712598.2025.2501731) carries four TILT Biotherapeutics affiliations - a competitor oncolytic-virus company. Named in A.5b rather than dropped, because 'the loudest advocate for this mechanism in this indication works for a competitor' is itself a finding.

Peak sales: a top-of-funnel epidemiology anchor now EXISTS where none did last time (~20,000 US and ~61,000 7MM incident ovarian cancer cases in 2024, ~80% recurrence after first line; WEB ESTIMATE, DelveInsight via PR Newswire, 2025-10-09), so the bottom-up attrition chain is written out. But TWO of its four material links carry no source at all - the fraction reaching platinum resistance and the fraction reaching a fourth line while still ECOG 0-1 - so no bottom-up point estimate is produced from it. The estimate still rests on the comparator route, now anchored on an ACTUAL rather than a forecast figure: Elahere global sales ~$750M in FY2025, up from ~$480M in 2024.

No composition-of-matter patent expiry could be verified on a SECOND attempt. The 2036 figure in a third-party database covers a combination patent on a different, out-licensed product (V2ACT Immunotherapy) and is not used.

No orphan-drug designation found on a second attempt; ChEMBL's orphan: 0 corroborates the absence in one more database without establishing the regulatory fact.

The Newsoara royalty rate is still not disclosed anywhere this sweep could read, on a second attempt, so the value of the Greater China rights is not estimated.

Competitor efficacy figures (ROSELLA, MIRASOL, KEYNOTE-B96, and the new ASCO 2026 Elahere platinum-sensitive failure) and the Elahere revenue anchor all come from secondary oncology-press reporting; primary publications were not opened. All tagged WEB ESTIMATE.

No quality-of-life or health-economic data exists for this agent in any source read.

Pre-existing anti-vaccinia immunity from prior smallpox vaccination is STILL not addressed in any source read, despite being the obvious intrinsic factor affecting exposure for a vaccinia-based agent.

First sweep of this ticker to include the three v5.6.0 mandatory company-tier rows (EDGAR submissions, EDGAR XBRL companyconcept, FINRA consolidatedShortInterest). All three returned, and two of them closed standing holes: the filed share count resolved the at-the-market drawdown question, and the FINRA count resolved the units of BPIQ's short_float field.
EDGAR companyconcept is the single most valuable new figure in this sweep. dei:EntityCommonStockSharesOutstanding is 44,840,416 as of 2026-05-03 and 44,805,811 as of 2026-03-16, a rise of only 34,605 shares across the seven weeks spanning the $100M ATM's establishment on 2026-03-19 - so the facility was essentially UNDRAWN through 2026-05-03. This was the largest single hole in the 2026-08-05 record. It does not cover 2026-05-03 to today.
EDGAR submissions establishes that Q2 2026 results had NOT been published as of 2026-08-11: the newest periodic filing is the Q1 10-Q of 2026-05-07, and the newest filings of any kind are three Form 4s dated 2026-08-04. Corroborated independently by the BPIQ press feed (date_updated 2026-08-10, no results release). The 2025 equivalent came 2025-08-07; the statutory deadline is 2026-08-14. Every cash, burn and runway figure therefore rests on a 2026-03-31 balance.
FINRA consolidatedShortInterest, settlement 2026-07-15: 3,113,408 shares short, average daily volume 174,004, days to cover 17.89, up 3.54% from 3,007,000 at 2026-06-30. Divided by the EDGAR share count that is 6.94% of shares outstanding and 7.99% of estimated float - close enough to BPIQ's short_float of 8.623% to establish that field's units as percentage-of-float. The 2026-07-31 settlement returned empty, which is the documented ~two-week publication lag, not a missing ticker.
fetch_company_info: both timestamps unchanged from the 2026-08-05 sweep. finance_updated_at 2026-03-31 (now four and a half months stale), short_data_updated_at 2026-07-15. The price feed still carries no date, but this sweep PINNED it: cross-checking two BPIQ sweeps against data/prices/GNLX.json shows last_price is the previous session's close, not an intraday quote, so $2.72 is the 2026-08-10 close and $2.66 the 2026-08-07 close.
market_cap bug proved exactly rather than inferred: 119,275,512 / 2.66 (previous close) = 44,840,418 against an EDGAR-filed 44,840,416, a match to two shares. max_52_week_position reproduces the same previous-close bug to five decimal places (reported 0.05925; (2.66-2.29)/(8.535-2.29) = 0.05925).
NEW connector inconsistency found: the connector's enterprise value cannot be reconciled to its own market cap and cash, and the residual is unstable - EV minus (market cap minus cash) is $2,494,994 today against $1,149,787 on 2026-08-05, with no financing or filing in between that could change a debt balance. The 2026-08-05 record reported that residual as approximately $1.15M of implied debt; this sweep declines to read it as debt at all.
fetch_company_drugs: 3 rows, unchanged in count, stage and catalyst text. All three are the same molecule, all three has_catalyst true, all three carrying the identical synthesized catalyst_date 2026-12-31 (one from 'H2 2026', two from '2026'). Drug names still dirty, one with a trailing space; keyed on the integer id throughout.
fetch_company_options_data: chain now dated 2026-08-10 and LESS readable than at the last sweep. Same four expiries and same three strikes ($2.50/$5.00/$7.50) on a $2.72 share, so still no at-the-money contract. Total open interest 527 contracts (up from 490) but day volume ZERO (down from 1). All four expiries now carry an IV artifact - 0.01488 floor on three $2.50 calls, each with delta exactly 1.0 and gamma exactly 0, and 9.99512 ceiling on the fourth - where at the last sweep one contract still produced an arithmetically real figure.
fetch_company_insider_transactions: 250 rows, now running to 2026-08-04. Exceeded the tool output limit again and was read from the saved file in full. Still no open-market purchase by any officer or director anywhere in the feed. CORRECTION FOUND: the 2026-06-16 grant block is 1,651,259 RSUs, not the 1,622,850 recorded on 2026-08-05 - the earlier figure counted 28,409 shares for three directors where the feed shows four (Thomas, Tyree, Smither, Mirabelli). Options unchanged at 126,264.
fetch_company_press_releases: 137 items, date_updated 2026-08-10, read in full. No new offering since 2026-03-19 and no Q2 results. Also carries no explanation for the 2026-08-05 price move.
UNEXPLAINED PRICE MOVE: data/prices/GNLX.json records a fall from $2.82 to $2.51 on 2026-08-05, about -11% on 949,100 shares against an average daily volume near 174,000. No company announcement appears in the press feed for that day or around it. Recorded, not explained. A third-party item dated 2026-08-06 independently describes the stock as 'near US$2.51'.
PRICE CACHE GAP: data/prices/GNLX.json was fetched 2026-08-06 and its final bar is 2026-08-05, five trading days before this sweep. The 52-week range is unaffected (neither extreme falls in the missing window) but the $2.72 spot rests on BPIQ alone and is not yet cross-checkable. Refreshing the cache is scripts/fetch-prices.mjs's job, outside this analysis and outside this lane.
fetch_company_hedge_fund_holdings: returned {} again - zero holders, corroborated by hf_holding_count 0 and in_no_hfs true. A legitimate empty still counts as CALLED.
fetch_company_earnings_transcript: returned no content on two further attempts. No transcript exists for this ticker through this connector. Optional, so it gates nothing.

Price cache

Derived from data/prices/GNLX.json — not a sweep row, so it carries no CALLED/BLOCKED state

First bar	2023-01-26
Last bar	2026-08-19
Bars	894
Fetched	2026-08-19

RARE Cleared

Company tier

Swept 2026-08-07 · called once, reused by every program

BPIQ:fetch_company_info	Called
BPIQ:fetch_company_drugs	Called
BPIQ:fetch_company_historical_catalysts	Called
BPIQ:fetch_company_options_data	Called
BPIQ:fetch_company_insider_transactions	Called
BPIQ:fetch_company_press_releases	Called
BPIQ:fetch_company_hedge_fund_holdings	Called
BPIQ:fetch_company_earnings_transcript	Called

Program tier — dtx301-otc-deficiency

BPIQ drug id 17190 · CLEARED

CT.gov:get_trial_details\|analyze_endpoints	Called
PubMed:search_articles+get_article_metadata	Called
Open Targets:search_entities	Blocked
ChEMBL:compound_search	Called
web_search:x4	Called

Open Targets BLOCKED for the whole session (4 attempts, verbatim rate-limit error) - the gene-disease corroboration is cited from the 2026-08-05 sweep's successful call and tagged as such; the association's numeric evidence score has never been retrieved in either sweep.

Still NO peer-reviewed publication of DTX301 human data: the name-keyed PubMed search returns 2 records (macaque study, sponsor-co-authored disease review), metadata retrieved on both; the previous sweep's 9-record count came from broader queries and the finding is identical.

The 64-week second primary endpoint has NO published success threshold - no definition, bar or statistical plan on the registry; the 'complete responder' definition used is Ultragenyx's consistent usage, UNVERIFIED as to protocol wording. This is why the probability band spans 50.

CATALYST-DATE DISAGREEMENT recorded UNRESOLVED per rule 24: company guidance H1 2027 (three dated statements: 2026-03-12, 2026-05-05, 2026-08-04) against CT.gov primary completion 2027-09, with the rule 34 modelled estimate (2027-11 to 2028-01) landing wholly after the guided half.

The BPIQ note field for id 17190 now carries only its newest entry, not the dated changelog it held on 2026-08-05 - the slip sequence's first two rows are carried from that sweep's verified reading; a count parsed from today's field alone would under-count.

CT.gov discloses no investigator names for NCT05345171 (no overall officials, no site contacts; search_investigators over 30 OTCD trials returned none for this NCT) - the KOL investigator table is empty over a fully-enrolled pivotal trial and the panel judgement is UNKNOWN.

After Week 36 the trial loses its contemporaneous control (crossover); the Week 64 analysis is a single-arm proportion. Every Week 36 efficacy figure comes from the company's own 2026-03-12 release (has_results false on the registry), re-read this session via a StockTitan rendering.

The previous analysis wrongly reported no Fast Track designation; the company's own release boilerplate states Fast Track (US) - corrected this version, sponsor-stated, no FDA-side confirmation pulled.

REGENXBIO NAV AAV8 royalty rate still unconfirmed; patent estate unresolved; orphan exclusivity (7y US, 10y EU) is the verified protection.

Anti-AAV8 seroprevalence (gates eligibility) still unquantified - a direct multiplier on every market figure.

No DTX301-specific peak-sales forecast exists; every figure is modelled. Analyst-target dispersion is extreme and partly stale ($34-$140, top predating the setrusumab crash; freshest actions Guggenheim $35 and Wells Fargo $45); no aggregate target used.

The committed price cache ends 2026-08-05 ($24.93) against the $26.67 quote read 2026-08-12; run-up exit resolves later against the cache and is null at lock.

Program tier — gtx-102-angelman-syndrome

BPIQ drug id 17188 · CLEARED

CT.gov:search_trials+get_trial_details	Called
CT.gov:search_investigators	Called
PubMed:search_articles+get_article_metadata	Called
Open Targets:search_entities Rate limit exceeded for client: global	Blocked
ChEMBL:compound_search	Called
web_search:x4	Called

ctgov_search_trials_and_get_trial_details: 5 studies on the intervention, 4 of them this program's; NCT06617429 in full, primary_completion_date now day-precise 2026-07-10; competitor trials NCT06914609 and NCT07605429 also pulled in full; NCT05531890 is a Grace Therapeutics name collision, excluded

ctgov_search_investigators: condition 'Angelman syndrome', 55 trials analysed, 66 investigators returned, NONE attached to NCT06617429

pubmed_search_articles_and_get_article_metadata: narrow query returns 2 records (up from 1), metadata on both; broad query returns 30, above the <=15 threshold, so metadata on 6 chosen for relevance - a disclosed narrowing; plus get_full_text_article on PMC13056972

open_targets_search_entities: retry policy exhausted (attempt, immediate retry, wait, retry) across ~40 minutes, four attempts. Unverified as a result: the catalogued UBE3A-Angelman association. The 2026-08-05 reading is cited as history, not as a live check. Nothing in the analysis depends on it - three sponsors in Phase 3 against the same transcript carries the target-validation score instead.

open_targets_search_entities: verbatim error — Rate limit exceeded for client: global

chembl_compound_search: legitimately empty for both 'apazunersen' and 'GTX-102', unchanged

web_search_x4: peak sales / competitive / exclusivity and royalty / analyst; royalty rate still not obtainable

fetch_company_info: finance fields still stamped 2026-03-31; short interest 2026-07-15; price fields are the 2026-08-06 close ($25.82)
fetch_company_drugs: nine rows, unchanged from the 2026-08-05 sweep; four has_catalyst true; 20225/20222 still a duplicate; four marketed products still absent
fetch_company_historical_catalysts: 22 rows, 2017-08-22 to 2026-04-02; no new row since the previous sweep
fetch_company_options_data: date_updated 2026-08-06; eight expiries; spreads widened to 52% of mid from 31%; single-strike day volume up from tens to thousands
fetch_company_insider_transactions: 250 rows, still ending 2026-07-01; two 2026-08-04 Form 4s appear in the press feed and in no row
fetch_company_press_releases: 406 items to 2026-08-05, nine more than the previous sweep; now reaches past the 2026-08-04 Q2 release
fetch_company_hedge_fund_holdings: optional; 18 quarters to 2026Q2; internally consistent, no 1000x bug
fetch_company_earnings_transcript: optional; still returns the Q1 2026 call (2026-05-05), three days after the Q2 call took place

Price cache

Derived from data/prices/RARE.json — not a sweep row, so it carries no CALLED/BLOCKED state

First bar	2021-08-06
Last bar	2026-09-18
Bars	1285
Fetched	2026-09-19

CGEM Cleared

Company tier

Swept 2026-08-10 · called once, reused by every program

BPIQ:fetch_company_info	Called
BPIQ:fetch_company_drugs	Called
BPIQ:fetch_company_historical_catalysts	Called
BPIQ:fetch_company_options_data	Called
BPIQ:fetch_company_insider_transactions	Called
BPIQ:fetch_company_press_releases	Called
BPIQ:fetch_company_hedge_fund_holdings	Called
BPIQ:fetch_company_earnings_transcript	Called
EDGAR:submissions	Called
EDGAR:xbrl_companyconcept_dei_EntityCommonStockSharesOutstanding	Called
FINRA:consolidatedShortInterest	Called
Yahoo:options_chain	Blocked

Program tier — zipalertinib-1l-egfr-ex20ins-nsclc

BPIQ drug id 17466 · CLEARED

CT.gov:search_trials+get_trial_details	Called
CT.gov:search_investigators	Called
PubMed:search_articles+get_article_metadata	Called
Open Targets:search_entities Rate limit exceeded for client: global	Blocked
ChEMBL:compound_search	Called
web_search:x4	Called
EDGAR:full_text_search	Called
EuropePMC:search	Called

Open Targets BLOCKED on FOUR attempts (initial, immediate retry, retry after several minutes of intervening work, narrowed single-ID attempt ENSG00000146648). No independent genetics-based validation of the EGFR-lung-cancer association; no evidence score asserted. Mitigated by three approved drugs against this exact mutation and independent academic crystallography (DOI 10.1073/pnas.2417144121). Third consecutive sweep of this ticker the platform throttle has blocked.

NO statistical assumptions for REZILIENT3 are public - the assumed hazard ratio, target event count, alpha allocation across the co-primary endpoints, existence of an interim efficacy analysis, and whether randomisation is stratified by insertion location were again not found in the registry, the protocol paper, or any search. The stratification question most affects the modelled probability and remains unanswered.

The catalyst date disagrees by FIVE MONTHS and is reported rather than reconciled per rule 24: partner guidance 'by the end of 2026' (stated identically three times: 2026-03-10, 2026-05-07, 2026-08-06) against a ClinicalTrials.gov primary completion of 2027-05-27. The registry lists primary completion and overall completion as the SAME date against outcome time frames of 'approximately 5 years', so the field is treated as unmaintained and the modelled estimate that keys on it as a late bound only.

has_results is false on NCT05973773; no REZILIENT3 efficacy data exist. Everything quantitative about zipalertinib comes from REZILIENT1 - a different trial, in a different line, with no control arm.

PubMed returned 27 records (>15); metadata retrieved on 14 (up from 10 on 2026-08-05). Thirteen older records unopened; nothing new on REZILIENT3 since the last sweep beyond an analytical-chemistry methods paper.

ChEMBL CHEMBL4650281 still carries oral: false for a twice-daily tablet - route flags wrong and unused - and returns no bioactivity data, so all selectivity claims rest on PubMed.

NO analyst peak-sales forecast specific to zipalertinib exists (re-verified). All B.3a figures modelled bottom-up; aggregate 'EGFR market' forecasts found and deliberately not passed through (rule 6).

The 50-60% pre-tax margin converting brand sales to Cullinan revenue is assumed, not disclosed by either party.

Patent estate and Breakthrough Therapy designation remain a WEB ESTIMATE from annual-report summaries; the filings were not opened, and the family count differs between report years (six per FY2023 vs seven per the FY2024 summary) - reported, not reconciled.

PAPILLON, sunvozertinib and furmonertinib figures come from 2026 peer-reviewed reviews and web search, not the primary trial publications.

The REZILIENT3 registry lists no overall officials or site investigators, so the KOL investigators panel is built from the programme's own publications; no conflicts search reached a journal disclosure appendix (PubMed metadata carries none; full texts not opened), and conflicts_checked records that limit rather than certifying clean records.

REZILIENT3 Part A's combination-safety conclusion rests on SIX patients.

The regulatory tier was NOT called: the catalyst is a trial data readout, and 02-connectors.md binds that tier only to regulatory events.

fetch_company_hedge_fund_holdings: NEW CONNECTOR LABELLING BUG FOUND. The filing_qtr label is the FILING quarter, not the holdings quarter: every value/shares ratio in the '2026Q2' block equals the committed price cache's 2026-03-31 close ($14.21), not the 2026-06-30 close ($18.21). The 2026-08-05 sweep took the labels at face value, so its implied quarter-end price series was mis-dated by one quarter throughout, including calling $5.93 'the 2025-12-31 trough' when it is the 2025-09-30 close. Corrected in this sweep's company.md C.4/C.5. Not documented in 02-connectors.md.
fetch_company_info: the cash field CHANGED COMPOSITION between sweeps - $75.081M on 2026-08-05 (narrow cash and equivalents, ~5x understatement of the disclosed $393.3M) vs $334.538M on 2026-08-10 (a broader line, ~6% understatement of the disclosed $356.0M). A field that silently changes composition cannot be trusted either way; the company's own Q2 2026 figure is used throughout.
fetch_company_info: enterprise_value ($807.898M) implies subtracting $281.75M of cash, matching neither its own cash field nor the disclosed total. Recomputed to $804.3M on the EDGAR-filed share count.
fetch_company_info: market_cap ($1,089.651M at $18.03) implies 60.4M shares against 64,354,694 filed at 2026-07-31. Recomputed from the EDGAR XBRL count - the v5.6.0 sweep row that exists for this bug.
fetch_company_info: tute_own (1.06111) and insider_own (0.00425) remain unreadable as fractions or percentages; both discarded. bpiq_pick_criteria is stale, still quoting the 2026-08-05 sweep's 52-week position and enterprise value.
fetch_company_drugs: 14 rows. SIX now carry has_catalyst true - velinotamig (19423, Q4 2026 SLE data) joined the catalyst list since 2026-08-05, making four disclosed Q4 2026 readouts on this one ticker plus CLN-978 RA in Q3.
fetch_company_historical_catalysts: 19 rows, 16 distinct dates (2026-06-10 x3, 2025-12-08 x2, identical price fields within each set).
fetch_company_options_data: response exceeded the tool output limit (50,757 chars) and was read in full from the saved file with jq. 308 contracts, 5,103 open interest, ZERO day volume chain-wide on the read date; near-money December spreads wider than their mids; 10 contracts on the 0.01488 IV floor artifact. The chain cannot price or bracket the event.
fetch_company_insider_transactions: response exceeded the tool output limit (57,256 chars) and was read in full with jq. 250 rows to 2026-07-06, nothing new since the 2026-08-05 sweep; the four-quarter limit again does not hold for this ticker.
EDGAR submissions (new v5.6.0 row): S-3ASR automatic shelf registration filed 2026-04-28 with no 424B5 following it - a live dilution mechanism on file where the 2026-08-05 sweep found none; new Schedule 13G by State Street Corp filed 2026-08-07; Form 4 feed matches the insider connector.
EDGAR XBRL companyconcept (new v5.6.0 row): 64,354,694 common at 2026-07-31; PreferredStockSharesOutstanding fell 481,989 -> 204,209 across Q2 2026, so the quarter's ~2.9M-share jump is preferred conversion (no proceeds), and the economic share count is now measured (~66.4M) rather than a floor.
FINRA consolidatedShortInterest (new v5.6.0 row): settlement 2026-07-15, 10,285,963 shares short, days-to-cover 15.22, ADV 675,823. 16.0% of filed shares outstanding; ~23% of the ~44M float, which corroborates BPIQ short_float's units (fraction of float) for this ticker for the first time. The 2026-07-31 settlement is not yet published - normal lag.
Yahoo options chain (optional): BLOCKED, verbatim 'Too Many Requests' from the crumb endpoint on two attempts. No coverage cost - optional row - but the BPIQ chain was the sole options source this sweep.
fetch_company_earnings_transcript: optional, returned no content for the second consecutive sweep.

Price cache

Derived from data/prices/CGEM.json — not a sweep row, so it carries no CALLED/BLOCKED state

First bar	2021-08-06
Last bar	2026-09-18
Bars	1285
Fetched	2026-09-19

TYRA Cleared

Company tier

Swept 2026-08-07 · called once, reused by every program

BPIQ:fetch_company_info	Called
BPIQ:fetch_company_drugs	Called
BPIQ:fetch_company_historical_catalysts	Called
BPIQ:fetch_company_options_data	Called
BPIQ:fetch_company_insider_transactions	Called
BPIQ:fetch_company_press_releases	Called
BPIQ:fetch_company_hedge_fund_holdings	Called
BPIQ:fetch_company_earnings_transcript	Called

Program tier — dabogratinib-ir-nmibc

BPIQ drug id 18677 · CLEARED

CT.gov:search_trials	Called
CT.gov:get_trial_details	Called
PubMed:search_articles	Called
PubMed:get_article_metadata	Called
OpenTargets:search_entities	Called
ChEMBL:compound_search	Called
web_search:peak_sales	Called
web_search:competitive	Called
web_search:exclusivity_royalty	Called
web_search:analyst	Called

ChEMBL holds no record of dabogratinib under three name variants (legitimately empty x3); selectivity ratios rest on the company's own peer-reviewed discovery paper with no independent replication.

CT.gov names no investigator or contact on any of SURF302's 47 locations; KOL investigator panel built from the SURF301 publication instead.

PMC full text of the dabogratinib clinical paper (PMC13060625) returned without its Declarations section, so its COI statement was not retrievable.

No public per-indication peak-sales figure found (analysts raised NMIBC models without publishing numbers); B.3a is a labelled model with the undefensible inputs named.

Patent estate read only through an aggregator (US 12,264,149 issued); not verified against the register.

The catalyst is a company-controlled interim: no registry field dates it; the window rests on twice-slipped company guidance plus enrollment arithmetic.

fetch_company_info financial fields one quarter stale (2026-03-31); Q2 release of 2026-08-04 supersedes cash ($353.9M) and runway ('into the second half of 2028'). tute_own reads 1.11832 (>100%) and is unusable; market_cap implies the stale fiscal-2025 share count (59.68M vs 65,991,171 Q2 weighted).
fetch_company_historical_catalysts intra_day_price_change_percent is wrong in sign on 2024-10-25 (+18.97% vs cache -23.3% close-to-close) and 2024-10-28 (+13.50% vs -10.8%); committed price cache used for all reaction figures.
Insider (250 rows to 2023-06-29) and press-release (236 items) responses exceeded tool output limits and were read in full from saved files. Press feed has a coverage hole: only one item predates 2024-10-28, so the February 2024 private placement is invisible to it.
fetch_company_earnings_transcript returned empty on two attempts - no transcript exists for TYRA via this connector; a legitimate empty result, recorded CALLED.
Options chain: 0.01488 IV floor artifact on August deep-ITM calls; $3-5 spreads on near-the-money strikes; readable for positioning (September $30/$40 call walls), not for an implied move.

Price cache

Derived from data/prices/TYRA.json — not a sweep row, so it carries no CALLED/BLOCKED state

First bar	2021-09-15
Last bar	2026-09-18
Bars	1258
Fetched	2026-09-19

VALN Cleared

Company tier

Swept 2026-08-11 · called once, reused by every program

BPIQ:fetch_company_info	Called
BPIQ:fetch_company_drugs	Called
BPIQ:fetch_company_historical_catalysts	Called
BPIQ:fetch_company_options_data	Called
BPIQ:fetch_company_insider_transactions	Called
BPIQ:fetch_company_press_releases	Called
BPIQ:fetch_company_hedge_fund_holdings	Not called
BPIQ:fetch_company_earnings_transcript	Not called
EDGAR:submissions	Called
EDGAR:companyconcept(dei:EntityCommonStockSharesOutstanding)	Called
FINRA:consolidatedShortInterest	Called
Yahoo:options_chain	Called

Program tier — s4v2-shigellosis

BPIQ drug id 18578 · CLEARED

CT.gov:search_trials+get_trial_details	Called
PubMed:search_articles+get_article_metadata	Called
Open Targets:search_entities Rate limit exceeded for client: global	Blocked
ChEMBL:compound_search	Called
web_search:x4	Called
CT.gov:search_investigators	Called

Open Targets search_entities BLOCKED across three attempts under the retry policy - the standing global throttle. Practical cost here is nil and is stated rather than assumed: Open Targets scores human gene-disease associations and this target is a bacterial surface polysaccharide, for which it holds no validation evidence.

ChEMBL count 0 on both 'Shigella4V' and 'S4V2' - a legitimate empty for a polysaccharide-protein bioconjugate. Off-target profile unaddressed by that connector; mechanism rests on the peer-reviewed structural characterisation (DOI 10.1093/glycob/cwz044) instead.

Neither S4V2 trial discloses a single investigator: all three US challenge-study locations and the one Kenyan infant-study site carry null contact fields, and no overall official is named. A condition-level search_investigators call over 50 shigellosis trials returned 26 names, NONE on NCT06615375 or NCT06523231. kol.investigators is an empty array with the search recorded.

No independent expert voice exists in this literature: all six indexed PubMed records carry LimmaTech authors, and every recurring academic name is a sponsor co-author. kol.independent_voices is empty with an UNKNOWN judgement. One name holds a dual relationship - Kawsar R. Talaat, first author of the Flexyn2a challenge paper with LimmaTech co-authors, is separately the CT.gov contact for EveliQure's competing ShigETEC Phase 2 challenge study (NCT07049159) at the same Johns Hopkins site listed on S4V03.

NCT06615375 primary_completion_date 2026-04-14 has PASSED with has_results false; NCT06523231 completed 2026-01-22 and has also posted no results. The press feed carries no S4V2 item since 2026-01-15. Central timing finding, not a gap.

S4V2 dose levels are not public (NCT06615375 says only 'high dose' and 'low dose'), so per-serotype antigen dose cannot be compared against Flexyn2a's 10 ug monovalent. What changed between S4V and S4V2 has never been disclosed either.

Peak-sales scenarios are TOP-DOWN, not bottom-up: no net price per dose is defensible for either the Gavi-tiered infant channel or the commercial traveller channel. Anchored on Valneva's own '>$500 million annually' market statement plus a share assumption, with the method named.

IP position could not be established: no patent number, claim type or expiry found for the S4V2 composition or the bioconjugation platform. The licence chain (GSK->LimmaTech July 2023, LimmaTech->Valneva August 2024) is verified; the GSK-to-LimmaTech financial terms are undisclosed, so the full royalty stack above Valneva's low-double-digit obligation is unknown.

attribution.status CLEAN is mechanical: VLA15 (13398) carries has_catalyst true with a null catalyst_date, so lib/clustering.mjs derives no window and drops it before any gap is measured. CLEAN means 'no other DATED catalyst is near this one', and both the attribution note and the stock-direction call say so.

First sweep of this ticker. fetch_company_drugs returned 7 rows; TWO carry has_catalyst true - 18578 (S4V2, this analysis) and 13398 (VLA15 Lyme), the latter with a NULL catalyst_date and catalyst_date_text 'TBA'.
The two BPIQ rows that returned empty on all three attempts are LEGITIMATE EMPTIES, both independently cross-checked, not the silent-failure pattern. fetch_company_options_data: VALN has NO exchange-listed options at all - an independent Yahoo options call returned expirationDates: [] and strikes: []. fetch_company_insider_transactions: Valneva is a foreign private issuer exempt from Section 16, and its entire EDGAR filing history contains no Form 3, 4 or 5.
fetch_company_historical_catalysts returned 14 rows but OMITS the two largest price reactions this equity has produced: the 2026-03-23 Lyme Phase 3 topline (-37.11% close-to-close) and the 2025-08-25 FDA IXCHIQ licence suspension (-18.99%). Both recovered from the press feed and priced from the committed cache.
intra_day_price_change_percent disagrees with the price cache on 2 of the 5 checkable rows (2025-11-04: BPIQ +2.66% vs cache -1.41% open-to-close; 2025-09-03: BPIQ +3.85% vs -0.79%); the other 3 match to the basis point. The 2025-09-03 row also carries news_published_at 2025-11-04 against catalyst_date 2025-09-03.
market_cap 541,234,112 cannot be reconciled with the filed share count: /$5.30 implies ~102.1M units against 189,771,237 ordinary shares (94,885,619 ADS-equivalents at the confirmed 1 ADS : 2 ordinary ratio). C.4 derives market cap (~$502.9M) from the filed count instead. The previous-close bug does not bite today (previous close = last price = 5.30).
cash and enterprise_value appear to carry EUR amounts with no currency label (cash 105,265,000 vs the disclosed EUR105.3M); the three burn fields are mutually inconsistent (ttm implies 1.57M/month vs monthly_burn 0.104M); cash_runway_end_date is the 2109-02-11 sentinel. None used - all recomputed or reported as absent.
fetch_company_press_releases returned 265 items to 2026-08-08 via the saved overflow file; read in full. NO S4V2 item since 2026-01-15.
Two optional BPIQ rows NOT CALLED (hedge_fund_holdings, earnings_transcript) - optional rows do not gate the sweep, and both questions were answered from the press feed and the primary releases instead.

Price cache

Derived from data/prices/VALN.json — not a sweep row, so it carries no CALLED/BLOCKED state

First bar	2021-08-11
Last bar	2026-08-19
Bars	1261
Fetched	2026-08-19

LGVN Cleared

Company tier

Swept 2026-08-11 · called once, reused by every program

BPIQ:fetch_company_info	Called
BPIQ:fetch_company_drugs	Called
BPIQ:fetch_company_historical_catalysts	Called
BPIQ:fetch_company_options_data	Called
BPIQ:fetch_company_insider_transactions	Called
BPIQ:fetch_company_press_releases	Called
BPIQ:fetch_company_hedge_fund_holdings	Called
BPIQ:fetch_company_earnings_transcript	Called
EDGAR:submissions	Called
EDGAR:xbrl_companyconcept	Called
FINRA:consolidatedShortInterest	Called
Yahoo:options_chain	Called

Program tier — laromestrocel-hlhs

BPIQ drug id 14275 · CLEARED

CT.gov:search_trials	Called
CT.gov:get_trial_details	Called
CT.gov:search_investigators	Called
PubMed:search_articles	Called
PubMed:get_article_metadata	Called
OpenTargets:search_entities Rate limit exceeded for client: global	Blocked
ChEMBL:compound_search	Called
web_search:peak_sales	Called
web_search:competitive	Called
web_search:exclusivity_royalty	Called
web_search:analyst	Called

Open Targets BLOCKED across three attempts (immediate retry, delayed retries >30 min apart) — no independent target-validation read; low cost for a multi-mechanism cell therapy but recorded rather than papered over

ChEMBL holds no record of a cell therapy (two searches, zero compounds): molecular identity and off-target profile not establishable from ChEMBL; identity rests on the WHO INN and trial records

The registered primary endpoint (CT.gov: RVEF at 12 months) and the approvable endpoint have diverged after the May 2026 Type C meeting; which endpoint the topline release leads with is unknowable in advance — the central interpretive risk in scoring this prediction

Two on-trial deaths disclosed but blinded (arm allocation unknown before topline)

CT.gov discloses no investigators for NCT04925024; the KOL panel is built from the program's publications instead

COI statements of the ELPIS I and CHILD papers not retrieved (metadata only); sponsor relationships read from author affiliation lists

fetch_company_options_data returned empty on three attempts; the Yahoo chain cross-check confirms LGVN has no listed options at all (expirationDates []), so the empty is legitimate — but the BPIQ empty is otherwise indistinguishable from a silent failure
fetch_company_hedge_fund_holdings filing-quarter label bug confirmed on LGVN: the '2026Q2' block's value/shares = $1.04 = the 2026-03-31 cache close, so it holds Q1 2026 holdings
BPIQ market_cap divides by the previous close (24,204,030 / 0.77 = 31.4M implied shares); recomputed from the 10-Q cover count (29,697,332 A + 1,449,005 B) x last price
BPIQ cash_runway_end_date 2027-02-18 conflicts with the company's own 'into Q4 2026' statement (Q1 2026 10-Q and call); both reported, neither picked
EDGAR XBRL: dei:EntityCommonStockSharesOutstanding absent for this dual-class filer (NoSuchKey); companyfacts WeightedAverageNumberOfSharesOutstandingBasic carries a x1000 scale artifact (CY2026Q1 reads 24,786,282,000); counts read from the 10-Q cover text
FINRA settlement 2026-07-15: 1,531,232 shares short, 4.8 days to cover; 2026-07-31 not yet published (normal lag)
The Q1 2026 earnings transcript, not any press release, carries the FDA's rejection of RVEF as the ELPIS II efficacy endpoint and the loss of the trial's pivotal label — a headline-only sweep would have missed the sweep's most important fact

Price cache

Derived from data/prices/LGVN.json — not a sweep row, so it carries no CALLED/BLOCKED state

First bar	2021-08-11
Last bar	2026-09-18
Bars	1282
Fetched	2026-09-19

CNTB Cleared

Company tier

Swept 2026-08-13 · called once, reused by every program

BPIQ:fetch_company_info	Called
BPIQ:fetch_company_drugs	Called
BPIQ:fetch_company_historical_catalysts	Called
BPIQ:fetch_company_options_data	Called
BPIQ:fetch_company_insider_transactions	Called
BPIQ:fetch_company_press_releases	Called
BPIQ:fetch_company_hedge_fund_holdings	Called
BPIQ:fetch_company_earnings_transcript	Called
EDGAR:submissions	Called
EDGAR:xbrl_companyconcept_dei_EntityCommonStockSharesOutstanding	Called
FINRA:consolidatedShortInterest	Called
Yahoo:options_chain	Blocked

Program tier — rademikibart-acute-asthma-exacerbation

BPIQ drug id 13433 · CLEARED

CT.gov:search_trials	Called
CT.gov:get_trial_details	Called
PubMed:search_articles	Called
PubMed:get_article_metadata	Called
OpenTargets:search_entities Rate limit exceeded for client: global	Blocked
ChEMBL:compound_search Internal error: Error calling tool 'compound_search': HTTP error in tool compound_search: 500 from upstream API (request ids 10f9b915-2726-46da-aa2a-0f7d8fa642ec, 455107e9-d602-41c7-81cb-05fde9e8442ec, 22e25699-ee50-4c4e-b1ee-a375068d4e24; three attempts, twice on 'rademikibart' and once on 'CBP-201')	Blocked
web_search:peak_sales	Called
web_search:competitive	Called
web_search:exclusivity_royalty	Called
web_search:analyst	Called
CT.gov:search_investigators	Called
EuropePMC:search	Not called
CTIS:search	Not called
EDGAR:full_text_search	Not called

PubMed get_article_metadata returned the AUTHORS FIELD AS NULLS on all 14 articles, while titles, journals, DOIs, dates and full abstracts all came back correctly. NOT PREVIOUSLY DOCUMENTED IN 02-connectors.md and the most consequential gap in this analysis: it is why kol.independent_voices is empty and kol.judgement.endpoint_supported is UNKNOWN, and why A.5's publication-quality row cannot assess author independence. The analysis rests instead on the abstracts themselves, which were complete.

Open Targets BLOCKED: no independent human-genetics validation of IL-4Ralpha. The target-validation claim rests on clinical evidence instead - dupilumab approved for eosinophilic asthma (2018) and COPD (2024) across eight indications, plus rademikibart's own positive published Phase 2b - and the genetics leg is tagged UNVERIFIED. Small loss on this particular target: an approved drug hitting the same receptor is stronger validation than a genetic association.

ChEMBL BLOCKED: no ChEMBL confirmation of molecule identity or off-target profile. Small for a structural reason - ChEMBL's selectivity data are built around small molecules and rademikibart is a monoclonal antibody, whose selectivity question is answered better by the peer-reviewed work in hand: a measured dissociation constant of 20.7 pM against IL-4Ralpha (vs dupilumab's 45.8 pM), no binding to IL-4Ralpha from other species, and a 2.71 A crystal structure of the Fab-receptor complex. The 500 is a server-side failure, not a rate limit, and is not the unreliable-indication-search caveat 02-connectors.md already records for ChEMBL.

NO PATENT OR EXCLUSIVITY INFORMATION obtainable. The dedicated exclusivity web search explicitly returned none for rademikibart's composition-of-matter claims or expiry, so B.1's IP row and B.2's exclusivity row are UNVERIFIED with no assumption substituted, and B.4 records the unknown as a risk rather than dismissing it.

CT.gov carries NO NAMED INVESTIGATOR for NCT06940141: all 58 sites return contacts null and there is no overall-official record. Normal once recruitment closes, and confirmed as a real empty rather than a tool failure by a second search_investigators call on the condition alone, which returned 21 investigators across 40 trials and none on a Connect Biopharma trial.

data/benchmarks/readout-lag.json holds ZERO OBSERVATIONS, so the modelled readout source uses the stated default of primary completion plus two to four months and is tagged 'UNVERIFIED - modelled, default lag'. On a trial with a 28-day primary endpoint whose safety window closed 2026-08-10 that default is probably too conservative; readout.basis says so rather than silently adopting it.

data/congresses.json holds no respiratory meeting at all, and separately the company has named no congress for this topline, so readout.sources congress is null with both reasons recorded. A respiratory congress inside the window is plausible given this company's history (EAACI 2025, ERS 2025, ATS 2025, AAD 2026) but matching on therapeutic area alone is a guess.

Europe PMC, CTIS and EDGAR full-text search NOT CALLED, all three optional. Europe PMC and EDGAR full-text search were skipped because PubMed returned the complete literature (14 of 14, total_count confirmed) and the EDGAR company feed was already read filing by filing at the company tier. CTIS was skipped because the trial's EU-region sites are in the United Kingdom, Serbia and Georgia - none in the CTIS jurisdiction - so a CTIS search would return a legitimate empty by construction.

Three of the four peak-sales inputs in B.3a are UNVERIFIED, and the least defensible is the PRICE: no drug is priced for a single acute dose in this setting anywhere in the world, so the $2,000-$6,000 per-dose range is an inference from a maintenance list price and an avoided-admission cost, not a benchmark. The eligible-event pool (roughly 1 million US asthma-related emergency-department visits a year, of which perhaps 30-50% type 2-high) is also unverified from a primary source.

The trial's eligibility criteria carry TWO eosinophil thresholds, both real: >=250 cells/uL historically (or FeNO >=25 ppb) when stable, and >=300 cells/uL at the index exacerbation. The analysis uses >=300, the enrolling threshold at the qualifying visit.

Source conflict left open: the company states 159 participants in the COPD study while CT.gov's NCT06940154 record says 160. Both recorded, neither picked. Does not affect this program.

The registry says Seabreeze STAT IV (NCT07705737) is NOT_YET_RECRUITING while the company says it initiated the study in August 2026. The registry lags.

Program tier — rademikibart-acute-copd-exacerbation

BPIQ drug id 18937 · CLEARED

CT.gov:search_trials	Called
CT.gov:get_trial_details	Called
PubMed:search_articles	Called
PubMed:get_article_metadata	Called
OpenTargets:search_entities Rate limit exceeded for client: global	Blocked
ChEMBL:compound_search	Called
web_search:peak_sales	Called
web_search:competitive	Called
web_search:exclusivity_royalty	Called
web_search:analyst	Called
CT.gov:search_investigators (optional, KOL sources)	Called

Open Targets is BLOCKED on all four attempts (query ['IL4R','chronic obstructive pulmonary disease','rademikibart'], then an immediate retry, then two further attempts spread across the sweep, per the 02-connectors.md retry policy). What is therefore unverified: independent human-genetics validation of IL-4Ralpha as a causal target in COPD. Cost is low but real - the target's relevance in this exact population rests on two approved drugs and a randomised placebo-controlled outcome (dupilumab's BOREAS and NOTUS), which is stronger evidence than a genetic association - so no claim in the analysis is tagged as genetically validated and A.5's Target validation row is scored on the trial evidence alone.

The DOSE used in Seabreeze STAT COPD is not disclosed by any source read this sweep. The CT.gov record lists the intervention as 'Rademikibart in prefilled syringe' with no dose, and no company release names one. This blocks any exposure-response reasoning about the acute regimen.

CT.gov names NO INVESTIGATOR for the pivotal trial. All 52 sites on NCT06940154 return contacts: null and the record names no overall official; search_investigators on condition 'COPD Acute Exacerbation' analysed 40 trials and returned 23 named people, none of them on NCT06940154. A sponsor choice about what to publish, not a connector failure - so kol.investigators is empty and there is no list to conflict-check.

PubMed returns NO CONFLICT-OF-INTEREST FIELD on any of the 14 articles. Every independence judgement in the kol block therefore rests on author affiliations, which the connector does return in full. Affiliation catches employment and trial-site relationships; it does not catch advisory-board seats, consulting arrangements or equity in a competitor.

IP POSITION UNEXAMINED. The dedicated exclusivity web search returned the Simcere Greater China licence terms but nothing on rademikibart's composition-of-matter claims or patent expiry. Not evidence of a problem - evidence that nobody has checked, on the company's only asset.

NO HEALTH-ECONOMIC MODEL or payer research exists for a per-event biologic in this setting. The payer argument in the analysis is an inference from avoided-admission costs rather than evidence, and B.3a's price input - the least defensible of the four peak-sales inputs - depends on it.

The eligible EVENT POOL for acute COPD exacerbations with eosinophils >=300 cells/uL (estimated 200,000-600,000 a year in the US) is derived rather than sourced: it is built from Sanofi's stated ~300,000 US adults addressable by dupilumab, widened for this trial's broader entry criteria and multiplied by an assumed one to two qualifying events per patient per year. Neither the broadening factor nor the events-per-patient rate was confirmed from a primary source.

First sweep of this ticker. Regulatory tier NOT CALLED AT ALL, correctly: 02-connectors.md makes it conditional on a regulatory catalyst and the only analysed catalyst here is a trial readout.
fetch_company_info market_cap ($146,706,000) is computed off a STALE SHARE COUNT, not a stale price: dividing by the last price $2.33 implies 62,963,853 shares, exactly the 2026-04-30 count. This is NOT the previous-close bug 02-connectors.md documents, which would have used $2.24. Recomputed on the EDGAR count of 62,974,309 at 2026-07-31 to $146.73M.
fetch_company_info enterprise_value ($90,031,283) implies subtracting $56.675M of cash, matching neither its own cash field ($27.015M), nor cash plus short-term investments ($31.486M), nor either at any prior quarter-end (2026-03-31 $46.034M; 2025-12-31 $44.342M). Unreconcilable; discarded and recomputed to $115.24M.
fetch_company_info max_52_week_position DID NOT reproduce the documented previous-close bug on this ticker: BPIQ's 0.42471 agrees with the LAST-price computation to five decimals, not with the previous-close computation of 0.390. The field is not read either way, but the bug is not universal.
fetch_company_info short_float (0.10992) is 3.4x the authoritative FINRA figure however read - 3.27% of shares outstanding or 4.80% of estimated non-affiliate float. Units unreadable; discarded in favour of FINRA divided by the EDGAR share count.
fetch_company_insider_transactions returned ZERO ROWS across two attempts while six Form 4s sit in the EDGAR feed over the same window. The bug 02-connectors.md records against TLSA, now REPRODUCED ON A SECOND TICKER, so it is a connector property rather than a TLSA quirk. The call completes rather than erroring, so the state is CALLED with an empty result; flow read from EDGAR Form 4 XML instead (director James Huang bought 1,160,000 shares at $3.45 on 2026-03-31 and 150,000 at $2.4774 on 2026-05-29, both open-market code P, both above the $2.33 spot).
fetch_company_earnings_transcript returned no output. Optional row, no consequence: the Q2 2026 release was read in full from the SEC as exhibit 99.1 to the 2026-08-12 Form 8-K.
fetch_company_hedge_fund_holdings: the CGEM filing-quarter bug DID NOT REPRODUCE AS DESCRIBED, and the labels could not be resolved to a quarter-end close by 02-connectors.md's prescribed method either. Dividing value by shares across the '2026Q2' block gives $2.619, $2.619, $2.619 and $2.434 - three agreeing and one not, so no single quarter-end close fits the block. Independently, ADAR1's $1,945,000 matches a holding publicly reported on 2026-07-28 as a 2026-06-30 position, which makes '2026Q2' the HOLDINGS quarter on this ticker rather than the filing quarter. The convention is therefore not consistent across tickers. Share counts used; dollar values not.
fetch_company_drugs: 9 rows, FOUR carrying has_catalyst true. Two of the four (19715, 19716) carry catalyst_date null and catalyst_date_text 'TBA' against Phase 1 IV data already reported on 2026-03-30 - stale flags on a delivered event. They contribute no window to the clustering computation, so they do not distort attribution, but a reader counting 'four catalysts' off the feed would be counting two that no longer exist.
fetch_company_drugs: drug names dirty as documented - five spellings of one molecule including 'Rademikibart /CBP-201 ' with a trailing space (id 13433). Every reference keys on the integer id.
fetch_company_options_data: 4 expiries, 3 strikes each (2.50/5.00/7.50) on a $2.33 stock, so no strike near spot and NO STRIKE BELOW SPOT AT ALL. 22 contracts of volume across the whole September expiry. The at-the-money put is quoted 0.55 bid against 3.50 ask, a six-fold spread; the same strike and expiry carries a 181% call IV against an 813% put IV. Both documented artifacts present and treated as null (9.99512 ceiling on the Sept $5.00 call; 0.01488 floor on three contracts). The chain cannot price or bracket the event.
EDGAR submissions (v5.6.0 row): TWO live registration statements that are different animals and were separated by reading both filing-fee exhibits. A $300,000,000 UNIVERSAL SHELF on Form F-3 filed 2025-06-10, POS AM 2025-09-10, effective 2025-09-12, carrying forward from 2022 file 333-264340 - real primary issuance capacity. AND a RESALE registration on Form F-3 filed 2026-05-15, effective 2026-06-01, for exactly 6,130,000 shares, which is the March 2026 private-placement share count - so it is supply overhang (9.7% of shares outstanding freed) rather than dilution. Whether the shelf's full amount is usable within twelve months was not settled: non-affiliate float sits near the $75M baby-shelf boundary but the affiliate share count was not confirmed.
EDGAR XBRL companyconcept (v5.6.0 row): 62,974,309 ordinary shares at 2026-07-31. Also used for cash ($27.015M), short-term investments ($4.471M), operating cash flow (-$31.927M for H1 2026) and liabilities ($16.611M total, no long-term debt concept reported at all).
FINRA consolidatedShortInterest (v5.6.0 row): settlement 2026-07-31, 2,060,037 shares short, days-to-cover 8.02, ADV 256,793. No publication-lag gap. Short interest NEARLY DOUBLED into the readout window - 1,061,303 (2026-06-15), 1,327,642 (06-30), 2,062,435 (07-15), 2,060,037 (07-31) - with the jump in the fortnight after the 2026-06-17 release that put a September date on the readout.
Yahoo options chain BLOCKED (verbatim: the crumb endpoint at query2.finance.yahoo.com/v1/test/getcrumb returned the body 'Too Many Requests' in place of a crumb, so the chain request was never made). Optional row. Consequence: the unusable BPIQ chain has no independent cross-check this sweep, and every C.6 figure is BPIQ-only. Does not change the conclusion that no source can price this event off this chain.
NO COMMITTED PRICE CACHE for this ticker - data/prices/CNTB.json does not exist. Per 02-connectors.md the price cache is not a coverage row and is derived from the cache files themselves, so it is recorded here as a note rather than a tool state. Two consequences carried into the analysis: 03a C.4's 52-week position was recomputed by hand under C.4's own stated exception, and the locked run-up call CANNOT BE SETTLED until the cache exists, because 05-prediction-protocol.md requires settlement to read the committed cache and never a fresh fetch. Fetching it is scripts/fetch-prices.mjs's job and writes outside analyses/CNTB/.
Source conflict left OPEN rather than smoothed: the company's filed 'cash runway expected for a period of at least one year from the date of this release' (2026-08-12) against this sweep's recomputation of 5.9 months from 2026-06-30 ($31.486M / $5.321M per month). Both recorded in company.md C.3, neither averaged nor picked. Both clear the September 2026 readout, so nothing in the analysis depends on resolving it.

Price cache

Derived from data/prices/CNTB.json — not a sweep row, so it carries no CALLED/BLOCKED state

First bar	2021-08-13
Last bar	2026-09-18
Bars	1280
Fetched	2026-09-19

MNOV Cleared

Company tier

Swept 2026-08-13 · called once, reused by every program

BPIQ:fetch_company_info	Called
BPIQ:fetch_company_drugs	Called
BPIQ:fetch_company_historical_catalysts	Called
BPIQ:fetch_company_options_data	Called
BPIQ:fetch_company_insider_transactions	Called
BPIQ:fetch_company_press_releases	Called
EDGAR:submissions	Called
EDGAR:xbrl_companyconcept_EntityCommonStockSharesOutstanding	Called
FINRA:consolidatedShortInterest	Called
BPIQ:fetch_company_hedge_fund_holdings	Not called
BPIQ:fetch_company_earnings_transcript	Not called
Yahoo:options_chain	Not called

Program tier — mn-001-nafld-hypertriglyceridemia

BPIQ drug id 16136 · CLEARED

CT.gov:search_trials	Called
CT.gov:get_trial_details	Called
PubMed:search_articles	Called
PubMed:get_article_metadata	Called
OpenTargets:search_entities Rate limit exceeded for client: global	Blocked
ChEMBL:compound_search Internal error: Error calling tool 'compound_search': HTTP error in tool compound_search: 500 from upstream API	Blocked
web_search:peak_sales	Called
web_search:competitive	Called
web_search:exclusivity_royalty	Called
web_search:analyst	Called
EDGAR:full_text_search	Not called
EuropePMC:search	Not called
CTIS:search	Not called
CT.gov:search_investigators	Called
Regulatory tier	NOT APPLICABLE

Open Targets BLOCKED after three attempts across roughly twenty minutes - the standing platform throttle 02-connectors.md already records. CONSEQUENCE: no independent human-genetic target validation for any of MN-001's targets against either co-primary endpoint. Every target-validation claim in the analysis is tagged UNVERIFIED and the analysis rests on the sponsor's own in vitro ABCA1/ABCG1/CD36 work plus a molecular-docking simulation instead.

ChEMBL BLOCKED after four attempts across roughly twenty minutes, HTTP 500 from the upstream API each time. CONSEQUENCE: the molecule's formal identity, structure and off-target selectivity profile are not independently confirmed; the mechanism description rests on published literature (PMID 17963431, PMID 40850750) and sponsor statements.

PubMed's index conflates 'MN-001' with the manganese crystal surface 'Mn(001)': three of seven hits are condensed-matter physics papers. A naive hit count would overstate this molecule's evidence base by more than double; the real figure is three non-bibliography papers in twenty years.

No peer-reviewed publication of MN-001-NATG-201 (NCT02681055) exists at all. The trial that is the entire prior evidence base for the pending readout is public only as registry entries and a conference poster whose three authors are all declared employees of the sponsor.

IRRECONCILABLE SPONSOR SOURCE CONFLICT on the prior primary endpoint: NATG-201's Week-8 triglyceride change reads -21.67 mg/dL (SD 27.89, n=19) in ClinicalTrials.gov's posted results and -138.8 mg/dL / -40.2% (345.7 -> 206.9, n=19) on the sponsor's own IDF 2022 poster - same endpoint, timepoint, denominator and units, differing by 6.4x. Reported per rule 24, not averaged and not resolved.

ClinicalTrials.gov NCT05464784 is stale: primary_completion_date and completion_date both still read 2026-12-31 eleven weeks after the sponsor announced last patient last visit on 2026-05-26. The registry field was therefore NOT used as the input to the modelled readout estimate, and the substitution is stated in readout.sources.

No investigator is named on the pivotal trial's registry record - two facilities, no contacts, no overall officials, no responsible-party investigator. kol.investigators is empty as a consequence. search_investigators over the condition returned one person, on an unrelated trial of a different company's drug.

No independent voice on either co-primary endpoint was found. Every named person attached to this program's data is a MediciNova employee. Two academic candidates (Ogura and Miida, Juntendo University) were excluded because a formal MediciNova-Juntendo MN-001 research collaboration was announced 2022-06-22, which is a disclosed sponsor relationship. kol.judgement.endpoint_supported is UNKNOWN as a result.

The Kyorin royalty rate is not disclosed anywhere - the 10-K says only 'a royalty on net sales of the licensed products'. No eNPV is written rather than one built on a guessed rate.

No defensible epidemiological count exists for the triple overlap (NAFLD + T2DM + triglycerides >150 mg/dL), so the base-case peak-sales patient count is named as the undefensible input. One third-party market figure ($12.82bn hypertriglyceridemia therapeutics market in 2026) is REJECTED under rule 6 as an aggregate including generic fibrates, statins and over-the-counter omega-3s.

No exposure-response analysis and no statistical analysis plan are public, so there is no disclosed multiplicity or alpha-allocation approach for two co-primary endpoints at 20 patients per arm.

The three optional NOT CALLED program-tier rows are deliberate: the 10-K was read directly for every MN-001 mention instead of EDGAR full-text search; Europe PMC's stated purpose is a fallback when a scholarly connector is BLOCKED and PubMed was not blocked; and both MN-001 trials in this indication are US-only, so an EU registry has nothing to add.

The regulatory tier is NOT APPLICABLE rather than NOT CALLED: 02-connectors.md binds it only to a regulatory event, and this catalyst is a trial readout.

First sweep of this ticker. No prior MNOV company document existed.
market_cap previous-close bug CONFIRMED on a second ticker after TLSA: BPIQ market_cap 64,972,048 divided by the EDGAR-filed 49,221,246 shares gives exactly $1.3200, the reported previous_close, not the $1.33 last_price. Recomputed at the last price: $65,464,257. max_52_week_position carries the same bug (0.18987 = ($1.32-$1.17)/($1.96-$1.17) exactly).
fetch_company_insider_transactions AGREES with the EDGAR filing feed on this ticker, unlike the TLSA case 02-connectors.md records: the connector's three June 2026 grants match three Form 4 accessions dated 2026-06-25, and the two January 2026 grants match two Form 4s dated 2026-01-20.
The options chain is unusable in a more basic way than the documented IV artifacts: the lowest listed strike on any of four expiries is $2.50 against a $1.33 spot, so there is no at-the-money strike and no straddle can be constructed. All ten put contracts additionally sit on the 0.01488 IV floor artifact with delta exactly -1. Total open interest across the whole chain is 78 calls and 10 puts.
FINRA short interest tripled in six weeks from a tiny base: 102,466 shares (2026-06-15) to 204,812 (2026-06-30) to 310,377 (2026-07-15) to 346,440 (2026-07-31); days-to-cover from 2.95 to 8.75. 0.704% of shares outstanding on the EDGAR count.
Cash and burn are 4.5 months stale (finance_updated_at 2026-03-31) and a Q2 2026 report previewed for 2026-08-08 was not filed at sweep time. insider_own and tute_own carry no timestamp at all.
The three optional NOT CALLED rows are deliberate and cost nothing against the gate: hf_holding_count is 0 so there are no 13F holders to enumerate, the 10-K and 10-Q were read directly instead of an earnings transcript, and the Yahoo options chain is a cross-check on IV artifacts that would not fix the absence of any strike near spot.
No price cache exists for this ticker (data/prices/MNOV.json absent), so C.4's 52-week range was computed by hand from BPIQ's own high and low per 03a-company-spec.md C.4's documented fallback. The price cache is not a coverage row and is derived at stage time from the cache files themselves.
BPIQ enterprise_value carries an unexplained $411,300 residual against market cap (previous close) less cash. Reported alongside the recomputed figure rather than smoothed over.

Price cache

Derived from data/prices/MNOV.json — not a sweep row, so it carries no CALLED/BLOCKED state

First bar	2021-08-13
Last bar	2026-08-19
Bars	1259
Fetched	2026-08-19

ARCT Cleared

Company tier

Swept 2026-08-13 · called once, reused by every program

BPIQ:fetch_company_info	Called
BPIQ:fetch_company_drugs	Called
BPIQ:fetch_company_historical_catalysts	Called
BPIQ:fetch_company_options_data	Called
BPIQ:fetch_company_insider_transactions	Called
BPIQ:fetch_company_press_releases	Called
EDGAR:submissions	Called
EDGAR:companyconcept:dei:EntityCommonStockSharesOutstanding	Called
FINRA:consolidatedShortInterest	Called
BPIQ:fetch_company_hedge_fund_holdings	Not called
BPIQ:fetch_company_earnings_transcript	Not called
Yahoo:options_chain	Not called

Program tier — arct-810-otc-deficiency

BPIQ drug id 16237 · CLEARED

CT.gov:search_trials	Called
CT.gov:get_trial_details	Called
CT.gov:search_investigators	Called
PubMed:search_articles	Called
PubMed:get_article_metadata	Called
OpenTargets:search_entities	Called
ChEMBL:compound_search	Called
web_search:peak_sales	Called
web_search:competitive	Called
web_search:exclusivity_royalty	Called
web_search:analyst	Called
EDGAR:full_text_search	Not called
EuropePMC:search	Not called
CTIS:search	Not called

No ARCT-810 clinical publication exists. Zero of 23 PubMed hits report ARCT-810 clinical data, and all four trials show has_results false. Every efficacy claim rests on company press releases; the analysis leans on trial-registry design facts and the committed price cache instead.

No efficacy threshold is pre-specified in the public record. The ureagenesis secondary endpoints read 'change from baseline' with no target magnitude, so the settlement definition is written around the regulatory consequence rather than a numeric bar.

No independent KOL voice could be recorded. The field's leading independent reviewers sit at UCL and Great Ormond Street Hospital, and Great Ormond Street was a site on Arcturus's own terminated Phase 2 (NCT05526066), which disqualifies them. kol.judgement.endpoint_supported is UNKNOWN.

Investigator conflicts could not be positively established: the only PubMed author record for Marshall Summar under urea-cycle terms is dated 2019, so any relationship formed since would not appear. conflicts is empty as a recorded search outcome, never as a clean bill of health.

The patent estate was not read. Exclusivity rests on the confirmed FDA Orphan Drug Designation (2019-06-27, seven years from approval); the combined term is UNVERIFIED.

CT.gov registry status for NCT06488313 is stale — RECRUITING with a primary completion date of 2026-06-01 already passed, against the company's 2026-08-06 statement that enrolment and dosing are complete. Only the status field is treated as out of date.

BPIQ carries no historical-catalyst row for the 2025-06-30 ARCT-810 interim data release, the most relevant precedent for this event. Reconstructed from the press feed and price cache (-2.4% close-to-close, after a +15.2% four-session run into the KOL-presentation announcement).

CT.gov search_investigators by investigator_name returned zero rows for an investigator the trial record names, because he is a site contact rather than an overall official. Queried by condition it returned him plus 66 mostly irrelevant people whose trials matched 'OTC' as optical coherence tomography, ovarian tissue cryopreservation and 'outside the cage' surgery.

Third-party market sizing rejected under rule 6: a $1,284M global OTC deficiency treatment market for 2025 is irreconcilable with a few thousand diagnosed patients and carries no derivation. B.3a is built bottom-up.

Aggregated analyst figures rejected in favour of named, dated ones: a web search returned a $68 high target and an H.C. Wainwright cut 'to $60 from $63', figures belonging to this stock's 2021-era price level rather than a $7.94 share.

ChEMBL returned a legitimate empty (count 0): ARCT-810 is an mRNA in a lipid nanoparticle with no ChEMBL entry. Cost is nil — off-target selectivity is not meaningful for an mRNA encoding a native human enzyme.

Two failure modes in 02-connectors.md did not fire here: ChEMBL did not return the HTTP 500 seen on CNTB and MNOV the same date, Open Targets did not return its standing global rate-limit block, and PubMed's authors-as-nulls bug did not fire.

market_cap fires the previous-close bug: BPIQ 216,015,312 ÷ EDGAR-filed 28,423,069 shares is exactly $7.6000, the reported previous_close, against a $7.94 last price. MNOV failure mode, not the CNTB stale-share-count one — the share count is current (2026-08-04).
max_52_week_position fires the same way: BPIQ 0.11248 = (7.60-5.50)/(24.17-5.50) exactly. Off the last price it is 0.1323. C.4 derives from the price cache and does not read the field.
enterprise_value does not reconcile against BPIQ's own cash field: 31,832,636 against a market-cap-less-cash of 24,532,312, an unexplained residual of 7,300,324. Debt is not separately disclosed, so it is reported as an open item rather than asserted as debt.
fetch_company_insider_transactions AGREED with the EDGAR Form 4 feed on this ticker — seven director grants dated 2026-06-05 match seven Form 4s filed 2026-06-15. The zero-rows gap recorded on TLSA and CNTB did not fire, confirming it is a per-ticker property rather than universal.
short_float AGREED with FINRA on this ticker: 7,206,773 short shares over an estimated non-affiliate float of 26,153,000 gives 27.56% against BPIQ's 0.27685, within 0.02 percentage points. The opposite of the CNTB disagreement.
FINRA's publication lag was shorter than 02-connectors.md records: that file notes the 2026-07-31 settlement unpublished as of 2026-08-10; it returned data on 2026-08-13.
An EDGAR submissions read returned another issuer's data mid-sweep (321 filings, 10-K dated 2026-03-18, 10-Q dated 2026-05-15 — none matching Arcturus). Caught because the wrong payload implied 424B5 offerings in Nov 2025 and Jan 2026 that the flat share count contradicted. Re-fetched and verified against the cik, name and tickers fields inside the payload, returning the correct 410-filing feed. Every EDGAR figure used comes from the verified re-fetch. A payload that parses cleanly and describes the wrong company announces nothing — verify identity inside the payload, not just the URL.
BPIQ intra_day_price_change_percent is open-to-close and hides overnight gaps. On 2025-10-22 it reports +16.57% for a session that was -50.2% close-to-close after a -57.3% gap. Every historical move used downstream was taken close-to-close from the committed price cache.
A press-feed headline dated 2026-03-23 reads 'Arcturus Therapeutics stock plunges 50% after weak cystic fibrosis trial data'. The price cache shows 2026-03-23 closed up 3.6%. The event it describes is 2025-10-22, five months earlier.

Price cache

Derived from data/prices/ARCT.json — not a sweep row, so it carries no CALLED/BLOCKED state

First bar	2021-08-13
Last bar	2026-08-19
Bars	1259
Fetched	2026-08-19

DWTX Cleared

Company tier

Swept 2026-08-13 · called once, reused by every program

BPIQ:fetch_company_info	Called
BPIQ:fetch_company_drugs	Called
BPIQ:fetch_company_historical_catalysts	Called
BPIQ:fetch_company_options_data	Called
BPIQ:fetch_company_insider_transactions	Called
BPIQ:fetch_company_press_releases	Called
BPIQ:fetch_company_hedge_fund_holdings	Called
BPIQ:fetch_company_earnings_transcript	Not called
EDGAR:submissions	Called
EDGAR:xbrl_companyconcept_dei_EntityCommonStockSharesOutstanding	Called
FINRA:consolidatedShortInterest	Called
Yahoo:options_chain	Blocked

Program tier — halneuron-cinp

BPIQ drug id 18554 · CLEARED

CT.gov:search_trials	Called
CT.gov:get_trial_details	Called
PubMed:search_articles	Called
PubMed:get_article_metadata	Called
OpenTargets:search_entities Rate limit exceeded for client: global (two attempts; the third, narrowed to a single query string, SUCCEEDED and returned SCN9A -> ENSG00000169432, so the row is CALLED)	Called
ChEMBL:compound_search	Called
web_search:peak_sales	Called
web_search:competitive	Called
web_search:exclusivity_royalty	Called
web_search:analyst	Called
CT.gov:search_investigators	Called
EuropePMC:search	Not called
CTIS:search	Not called
EDGAR:full_text_search	Not called

OpenTargets recovered on the third attempt, and that is the finding worth carrying forward: a run that gave up after two attempts would have recorded BLOCKED where data was actually available. Narrowing the query from three strings to one is what got through.

ChEMBL compound_search returned the molecule (CHEMBL507974, max_phase 3, natural product, MW 319.27, synonyms including 'Tectin' - which independently confirms Halneuron and Wex's Tectin are one molecule) but NOT a cross-subtype bioactivity panel. The claim that tetrodotoxin blocks six of the nine Nav subtypes is therefore tagged as established background rather than as a verified measurement from this session. The HTTP 500 that blocked this tool on both tickers swept earlier the same day did NOT recur.

PubMed's AUTHORS FIELD CAME BACK COMPLETE on all 7 fetched articles - a negative finding against the CNTB null-authors bug documented the same day. Recorded because it is what made the A.5b independence assessment possible at all: four of the six authors on the pivotal predecessor publication were employees of the sponsor or its CRO, and a run that assumed the bug would recur would never have checked.

THE PIVOTAL TRIAL NAMES NO INVESTIGATOR at any of its 25 sites. Every location on NCT06848348 is registered as 'Central Recruiting Site' with no facility name and no contact, there are no overall officials, and the only listed contact is a company email (Mehran@dwtx.com). CT.gov search_investigators returned zero. The predecessor trial NCT01655823 named all 23 of its sites including UT Southwestern and the John Theurer Cancer Center. Unexplained, and the reason kol.investigators is empty. kol.independent_voices is empty for a different and structural reason: the complete indexed literature on this molecule in pain is seven papers, every one generated by the sponsor's own program.

ClinicalTrials.gov NCT06848348 is STALE - last updated 2026-05-29 with an ESTIMATED primary completion of 2026-07 that has already passed. This is why readout.precision is PERIOD rather than MONTH, and why the run-up priority score is capped at 15 by the date-confidence gate.

PubMed's broad query returned 213 hits, over the <=15 threshold for fetching every hit. Narrowed to 'tetrodotoxin[Title] AND (pain OR neuropathy) AND (Clinical Trial[Publication Type] OR Randomized Controlled Trial[Publication Type])', which returned exactly 7 - the complete clinical-trial literature for this molecule - and all 7 were fetched.

First sweep of this ticker. Regulatory tier NOT CALLED AT ALL, correctly: 02-connectors.md makes it conditional on a regulatory catalyst and the only catalyst on this pipeline is a trial readout.
BPIQ:fetch_company_earnings_transcript is NOT CALLED and that is deliberate on an optional row: the Q2 2026 numbers were read in full from the primary source, Form 8-K exhibit 99.1 filed 2026-08-13 at 08:50, which is better than a transcript.
fetch_company_options_data returned NO OUTPUT across three attempts (immediate retry then after an interval). NOT PREVIOUSLY DOCUMENTED in 02-connectors.md for this connector. The tool's contract returns None for both failure and emptiness, so an empty chain and a silent failure are indistinguishable and no verbatim error exists to record - hence CALLED with an empty result rather than BLOCKED, the same reading 02-connectors.md gives the insider connector's zero-row case. The Yahoo cross-check that would have settled which it was was itself blocked, so company.md C.6 records the chain as unreadable from two independent directions and no options-implied move is used anywhere in the analysis.
fetch_company_insider_transactions returned ZERO ROWS across three attempts while thirteen Form 4s sit in the EDGAR feed over the same window. THIRD TICKER on which this is now recorded, after TLSA and CNTB, so the connector property is confirmed rather than suspected. Flow read from EDGAR Form 4 XML instead: routine option grants only (four directors 16,800 each on 2026-06-16; 330,000 to CEO Gregory Duncan on 2026-03-05), and NOT ONE open-market purchase or sale against a stock that fell from $9.50 to $1.28 over that window.
market_cap ($64,736,884) reproduces the PREVIOUS-CLOSE bug cleanly: divided by the EDGAR-filed 33,629,553 shares it is exactly $1.9250, the reported previous close, against a $1.9346 last close. Recomputed to $65.06M.
enterprise_value ($48,459,172) is UNRECONCILABLE against every available cash figure: it implies subtracting $16,277,712, matching neither the actual $9,610,033 at 2026-06-30 nor BPIQ's own stale $13,227,839 at 2026-03-31. Discarded and recomputed to $55.45M.
cash is ONE QUARTER STALE (finance_updated_at 2026-03-31) while the 2026-06-30 figure was public from the morning of the sweep.
max_52_week_position DID reproduce the previous-close bug here (0.07847 = (1.925-1.28)/(9.50-1.28)), where it did NOT reproduce on CNTB eight hours earlier. Second documented outcome for a bug that fires on some tickers and not others. The field is not read either way; C.4 derives the position from the price cache.
short_float (0.01398) AGREES with the FINRA-derived 1.45% of non-affiliate float on this ticker - against CNTB the same day where it was 3.4x the FINRA reading under every interpretation. The field is unreliable, not reliably wrong.
fetch_company_press_releases' newest item is 2026-07-20 and MISSES the Q2 earnings release filed the morning of the sweep. A press feed is not a substitute for the EDGAR filing index on the day a company reports.

Price cache

Derived from data/prices/DWTX.json — not a sweep row, so it carries no CALLED/BLOCKED state

First bar	2021-08-13
Last bar	2026-08-19
Bars	1259
Fetched	2026-08-19

EVMN Cleared

Company tier

Swept 2026-08-13 · called once, reused by every program

BPIQ:fetch_company_info	Called
BPIQ:fetch_company_drugs	Called
BPIQ:fetch_company_historical_catalysts	Called
BPIQ:fetch_company_options_data	Called
BPIQ:fetch_company_insider_transactions	Called
BPIQ:fetch_company_press_releases	Called
EDGAR:submissions	Called
EDGAR:xbrl_companyconcept	Called
FINRA:consolidatedShortInterest	Called
BPIQ:fetch_company_hedge_fund_holdings	Not called
BPIQ:fetch_company_earnings_transcript	Not called
Yahoo:options_chain	Not called

Program tier — evo756-atopic-dermatitis

BPIQ drug id 20071 · CLEARED

CT.gov:search_trials	Called
CT.gov:get_trial_details	Called
PubMed:search_articles	Called
PubMed:get_article_metadata	Called
OpenTargets:search_entities	Called
ChEMBL:compound_search	Called
web_search:peak_sales	Called
web_search:competitive	Called
web_search:exclusivity_royalty	Called
web_search:analyst	Called
CT.gov:search_investigators	Called

ChEMBL holds no record of EVO756 (legitimate empty, count 0 - NOT the HTTP 500 recorded on other tickers), so there is no independent selectivity or off-target profile. Every selectivity claim rests on the company's own 'highly selective' description, which is UNVERIFIED. Cost is low for the safety question (three clean human trials are better evidence) and REAL for the question of whether the CSU failure reflects insufficient target coverage rather than a wrong target, which no source this sweep reached can answer.

NO peer-reviewed publication of EVO756 clinical data exists anywhere. PubMed 'EVO756' returns exactly two records: a 2026 MRGPRX2 review reporting no EVO756 data, and a 2009 hybrid-speciation paper on butterflies whose publisher item identifier is literally the string 'EVO756' - the documented short-code collision in its purest observed form. Every efficacy and safety claim traces to a company press release.

The Incyte EP262/INCB000262 toxicology pause and halt could NOT be verified from a primary source: the Fierce Biotech article returned HTTP 403 Forbidden to a direct fetch, so it is tagged WEB ESTIMATE throughout. Nothing in the prediction depends on it - it strengthens EVO756's relative safety position and weakens the class's clinical validation, and those roughly cancel.

The composition-of-matter patent term for EVO756 could not be established, and the upstream royalty rate owed to Dermira (now Eli Lilly) under the December 2020 worldwide in-licence is not public. Both recorded as open gaps rather than estimated; they are why every eNPV figure is a range. Neither touches the September readout.

NO named investigators exist to assess: CT.gov search_investigators returned zero, corroborated by get_trial_details, which lists 29 sites with no overall officials and no site contacts. Normal for a commercially run dermatology Phase 2.

NO qualifying independent KOL voice was found. Every named MRGPRX2 commentator identified was disqualified for a stated reason - Escient/Incyte employment (Wollam, Viswanath, Boehm), co-authorship on the competitor-sponsored paper (Metz, Frischbutter, Gaudenzio), a competing academic compound plus mouse-only data (Kumar, Chow), or no comment on this program at all (Roth, Dong, Kim, Ali). kol.judgement.endpoint_supported is UNKNOWN as a result. NOTE that the PubMed null-authors bug did NOT fire this sweep - all eight articles returned complete author lists, which is the only reason this assessment was possible.

The three NOT CALLED rows are all OPTIONAL rows in 02-connectors.md's company tier and carry no gate consequence; they are recorded so the omissions are visible rather than silent. Fund positioning was read from the press feed's 13F alerts and BPIQ's own hf_holding_count; the Q2 2026 release was read in full from the company's IR site, a stronger source than a transcript; and the Yahoo chain would not have helped because C.6's problem is absent bids and absent open interest, which a second quote source cannot manufacture.
fetch_company_insider_transactions returned ZERO ROWS across THREE attempts with no error text, against 64 Form 3/4 filings in the EDGAR feed. The TLSA/CNTB bug on a third ticker, so it is firmly a connector property. State is CALLED with an empty result. Cost was material: the two July 2026 open-market purchases - the only discretionary insider trades on the record - are invisible to the connector entirely.
cash UNDERSTATES the balance sheet by $87.2M - a NEW failure mode for this field in this corpus. BPIQ's $200,793,000 is cash plus SHORT-TERM investments only and omits $87.214M of long-term investments the company itself counts when it states '$288.0 million'. enterprise_value inherits the error and is internally consistent with it (residual $143,579), so that field is not independently broken.
market_cap and max_52_week_position BOTH fire the documented previous-close bug cleanly on this ticker, unlike CNTB where max_52_week_position did not fire at all. $512,444,608 / 36,292,113 shares = exactly $14.1200; (14.12-10.47)/(33.20-10.47) = 0.16058 to five decimals.
monthly_burn shows NO bug: BPIQ's $6,772,000 is exactly the filed Q2 2026 operating cash outflow of $20.316M divided by three, and qtr_burn matches the filed figure to the dollar. Recorded so a reader does not assume every BPIQ field here was rejected.
tute_own (1.15620) is unreadable in either unit - 115.6% is impossible, 1.16% absurd for a name eight funds hold. Discarded; the institutional and retail ownership rows are left explicitly unfilled.
short_float (0.23264) is 1.88x the FINRA-over-shares-outstanding figure of 12.35%. UNLIKE CNTB the gap has a plausible innocent explanation (it would be right if the field means percent of float and the float is 53% of the share count), but that is an inference, not a confirmation, and the field is not used.
Three pipeline rows carry the IDENTICAL drug_name string 'EVO756 (MRGPRX2 Antagonist)' for three different indications (20071 live in AD, 20079 discontinued in CSU, 20081 in migraine). The name does not distinguish the live program from the failed one; indications_text separates them and the integer id is what everything keys on.
Row 20071's catalyst_date_text ('Q3 2026') is STALE RELATIVE TO BPIQ'S OWN note FIELD on the same row, which records the company's 2026-08-06 update to 'September 2026'.

Price cache

Derived from data/prices/EVMN.json — not a sweep row, so it carries no CALLED/BLOCKED state

First bar	2025-11-06
Last bar	2026-09-18
Bars	217
Fetched	2026-09-19

IRON Cleared

Company tier

Swept 2026-08-20 · called once, reused by every program

BPIQ:fetch_company_info	Called
BPIQ:fetch_company_drugs	Called
BPIQ:fetch_company_historical_catalysts	Called
BPIQ:fetch_company_options_data	Called
BPIQ:fetch_company_insider_transactions	Called
BPIQ:fetch_company_press_releases	Called
EDGAR:submissions	Called
EDGAR:xbrl_companyconcept	Called
FINRA:consolidatedShortInterest	Called
BPIQ:fetch_company_hedge_fund_holdings	Not called
BPIQ:fetch_company_earnings_transcript	Not called
Yahoo:options_chain	Not called

Program tier — disc-3405-polycythemia-vera

BPIQ drug id 18007 · CLEARED

CTgov:search_trials	Called
CTgov:get_trial_details	Called
CTgov:search_investigators	Called
PubMed:search_articles	Called
PubMed:get_article_metadata	Called
OpenTargets:search_entities	Called
OpenTargets:query_graphql Rate limit exceeded for client: global	Blocked
ChEMBL:compound_search	Called
web_search:peak_sales	Called
web_search:competitive	Called
web_search:exclusivity_royalty	Called
web_search:analyst	Called
EDGAR:full_text_search	Not called
EuropePMC:search	Not called
CTIS:search	Not called

No independent human-genetics ASSOCIATION evidence for TMPRSS6 in polycythemia vera. Open Targets' search layer answered this sweep and resolved TMPRSS6 to ENSG00000187045, polycythemia vera to MONDO_0009891 and HAMP to ENSG00000105697 - so the mandatory search_entities row is CALLED, an improvement on the 2026-08-13 sweep which recorded it BLOCKED - but three attempts at the follow-up association query target(ensemblId:'ENSG00000187045'){associatedDiseases{...}} each returned the verbatim error above. Resolving an identifier is not validating a target. The A.1 target-validation claim therefore still rests on the Mendelian phenotype (iron-refractory iron-deficiency anaemia) established in the clinical genetics literature and is tagged UNVERIFIED accordingly.

No investigator on RESTORE-PV could be named. get_trial_details on NCT06985147 returns all 15 facilities with contacts: null and no overall-official record, the normal state for an ACTIVE_NOT_RECRUITING trial. Confirmed as a real empty by search_investigators on 'Polycythemia Vera' across 40 trials, which returned 72 named investigators and contacts and not one on a Disc Medicine trial - a completely different result set from the 35 returned on 2026-08-13, which shows the connector sampling live trials rather than failing. kol.investigators is empty as a recorded finding.

No independent voice could be recorded. The only named panel taking a substantive position on this endpoint in this disease (Kremyanskaya, Ginzburg, Hoffman, Mount Sinai) authored a review advocating for a direct competitor by name, and their conflict-of-interest statements could not be obtained: PubMed get_article_metadata returns no conflict field on any of the eight articles retrieved and both journals are paywalled. kol.judgement.endpoint_supported is consequently UNKNOWN.

No defensible epidemiology figure for the phlebotomy-dependent polycythemia vera population in the launch markets was obtained. The peak-sales build routes around it through a third-party market projection whose methodology is not published, with share assumptions that have no external anchor at all. This is the weakest input in the analysis and it is named as such rather than filled.

ChEMBL compound_search returned a legitimate empty (count 0, total 0) on both 'DISC-3405' and 'MWTX-003', as on the previous sweep. ChEMBL indexes small molecules and this is a humanized monoclonal antibody, so the selectivity question it would answer is answered instead by the binding target and the immunogenicity data in the Phase 1 publication. No HTTP 500 on either call, so the failure mode recorded on CNTB and MNOV on 2026-08-13 did not recur.

Whether Mabwell's 9MW3011 (NCT06752746, anti-TMPRSS6 antibody in polycythemia vera in China) is the same molecule as DISC-3405 could not be established this session either. Recorded as an open question rather than resolved by assumption.

Full company-tier re-sweep on 2026-08-20, replacing the 2026-08-13 sweep. The caller asked for a Company refresh explicitly, so the tier was re-run in full regardless of what 02-connectors.md § Company-sweep currency would otherwise have permitted. No mandatory row reads NOT CALLED; the three NOT CALLED rows are optional by 02-connectors.md's own table and do not bear on the gate.
fetch_company_insider_transactions returned ZERO ROWS ON THREE ATTEMPTS with no error - the documented connector bug, reproduced on this ticker for a second consecutive sweep and on a fourth ticker overall after TLSA and CNTB. 36 Form 4 filings sit in the 2026 EDGAR feed, three of them filed inside the 2026-08-13 to 2026-08-20 delta window. Insider flow was read from EDGAR and the three new Form 4s were parsed individually from ownership.xml.
THE COMMITTED PRICE CACHE CARRIES THREE DEFECTIVE BARS ON THIS TICKER, and one of them was locked against by the superseded 2026-08-13 record. scripts/fetch-prices.mjs never overwrites a non-null value, so a bar first written mid-session stays an intraday snapshot for ever: 2026-08-13 is stored as $79.50 on 66,796 shares against a discarded $80.01 on 341,000; 2026-08-19 as $80.58 on 79,634 against a discarded $80.32 on 457,200; and 2026-08-20 is a partial bar by construction at $79.51 on 106,841. The 2026-08-19 case is independently confirmed by BPIQ's own previous-close field for 2026-08-20, which reads $80.32. This lock's spot therefore carries a disclosed 0.7% uncertainty (BPIQ read $78.92 the same day) and is labelled as an intraday reading rather than a close throughout the documents. The cache is not a coverage row and this note records the finding rather than a tool state.
max_52_week_position and market_cap both fire the documented previous-close bug for a second consecutive sweep: BPIQ's 0.67765 and $3,083,713,280 both resolve exactly on its own $80.32 previous close against the EDGAR-filed 38,392,844 shares. Both recomputed from the price cache and the filed share count in company.md C.4.
enterprise_value has become NEARLY RECONCILABLE where it was not. On 2026-08-13 BPIQ's figure implied subtracting $745,571,404 from its own market cap, matching no balance-sheet figure; this sweep it implies subtracting $657,267,979 against an actual net cash position of $658,406,000 - a $1.14M (0.05%) residual where a week ago there was an $87M hole. The likeliest reading is that BPIQ picked up the June 2026 debt drawdown, but that is an inference. The field is still wrong on its price basis, so it is recomputed rather than quoted.
tute_own remains unreadable and was discarded (1.14916 this sweep against 1.14464 a week ago - live and still uninterpretable). The institutional-ownership row records a gap rather than picking a reading.
FINRA's newest published settlement is still 2026-07-31, so short interest is identical to the previous sweep. The 2026-08-14 settlement had not published by this sweep date, which is ordinary publication lag rather than a missing ticker; short interest through the first half of August is unobservable and is recorded as a stated blind spot.
The press feed still carries two misdated MSN aggregator copies of the October 2025 offering as 2026 events (2026-03-20 and 2026-04-26), and adds a recurring Rosen Law Firm investor solicitation (2026-08-15, the sixth since 2026-04-03). Both are noise: the 2026 EDGAR feed contains no registration statement of any kind, and the Form 10-Q filed 2026-07-30 states verbatim that the company faces no material legal proceedings pending or threatened.
fetch_company_historical_catalysts returned the same thirteen rows as the previous sweep, newest 2026-06-12; no new catalyst reaction to record. The 2026-02-13 Complete Response Letter - the largest single-day move in this company's history - remains absent from that feed entirely and was again sourced from the press feed and the price cache.

Price cache

Derived from data/prices/IRON.json — not a sweep row, so it carries no CALLED/BLOCKED state

First bar	2021-08-13
Last bar	2026-09-18
Bars	1280
Fetched	2026-09-19

CUE Cleared

Company tier

Swept 2026-08-19 · called once, reused by every program

BPIQ:fetch_company_info	Called
BPIQ:fetch_company_drugs	Called
BPIQ:fetch_company_historical_catalysts	Called
BPIQ:fetch_company_options_data	Called
BPIQ:fetch_company_insider_transactions	Called
BPIQ:fetch_company_press_releases	Called
BPIQ:fetch_company_hedge_fund_holdings	Called
BPIQ:fetch_company_earnings_transcript	Not called
EDGAR:submissions	Called
EDGAR:companyconcept dei:EntityCommonStockSharesOutstanding	Called
FINRA:consolidatedShortInterest	Called
Yahoo:options chain	Not called

Program tier — cue-221-csu

BPIQ drug id 20405 · CLEARED

CT.gov:search_trials	Called
CT.gov:search_by_sponsor	Called
CT.gov:get_trial_details	Called
CT.gov:analyze_endpoints	Called
CT.gov:search_investigators	Called
PubMed:search_articles	Called
PubMed:get_article_metadata	Called
Open Targets:search_entities Rate limit exceeded for client: global	Blocked
ChEMBL:compound_search	Called
web_search:peak sales	Called
web_search:competitive	Called
web_search:exclusivity + royalty	Called
web_search:analyst	Called
EDGAR:full-text search	Not called
Europe PMC:search	Not called
CTIS:search	Not called

THE PIVOTAL TRIAL HAS NO ClinicalTrials.gov RECORD. Confirmed by three independent search strategies on 2026-08-19: by condition and intervention (15 hits, none of them it), by sponsor 'Ascendant' (0 hits), and by condition + China + Phase 2 (14 hits, none of them it). The three registered UB-221 trials (NCT03632291, NCT04175704, NCT05298215) are all earlier Taiwan studies and none produces this catalyst. Consequence: readout.sources[].ctgov is null with a note, the modelled estimate cannot use the standard primary-completion-plus-lag arithmetic, and the trial's size, arms, comparator, primary endpoint and analysis timepoint are unverifiable before the result lands. The connector was called and answered honestly - this is a data gap, not a coverage gap, which is why the program is CLEARED rather than PROVISIONAL.

Open Targets BLOCKED on the standing global rate limit across three attempts (immediate, retried, and again later in the sweep). No independent genetic validation of the IgE / FcepsilonRI / CD23 axis. The cost is small and is stated in the document: target validation here rests on omalizumab's 2014 approval against the same target in the same disease, which is stronger evidence than any genetic association, and no claim in the analysis depends on Open Targets.

ChEMBL compound_search on 'UB-221' returned a legitimate empty result ({"count":0,"total":0,"compounds":[]}) while the same tool returned a full record for 'omalizumab' (CHEMBL1201589) on the same sweep - an absence in ChEMBL's coverage, not a failure. No selectivity or off-target data, of limited consequence for a monoclonal antibody.

Two of the three registered CUE-221 trials sit at status UNKNOWN, years past their primary completion dates, with no results posted: NCT05298215 (primary completion 2023-12-31, the only registered Phase 2) and NCT04175704 (primary completion 2024-11-30). No efficacy figure for this molecule has ever been published beyond the single-dose findings in the 2022 J Clin Invest paper.

The evidence base is one primary research article. PubMed 'UB-221' returned 9 hits, all 9 retrieved and read by title: one primary paper (PMID 35912861), one letter, seven reviews. 27 of the primary paper's 29 authors are employed by United BioPharma or its affiliates and the differentiating CD23 finding has never been independently replicated. The PubMed authors-returns-nulls bug recorded in 02-connectors.md did NOT fire on this sweep - complete author lists came back on all nine.

No conflict-of-interest statement could be read for any person named in the kol block. PubMed get_article_metadata returns metadata and abstracts but not declaration-of-interest sections, so every conflicts_checked entry certifies that the search was made and states its own limit rather than certifying an absence. Marcus Maurer, the most-cited commentator in this indication, was deliberately NOT recorded as an independent voice because his disclosure set across competing CSU sponsors was not verified this sweep.

Published analyst targets are unusable and were rejected under 01-rules.md rule 6 rather than passed through: aggregators mix pre- and post-1-for-30-reverse-split figures, giving a $3.00 to $115.00 spread that cannot describe one security. No analyst target is used as an anchor anywhere in the analysis.

Five pipeline rows returned; exactly one (id 20405) carries has_catalyst true, which is why the clustering pass reads CLEAN for it.
market_cap is computed off the previous close: $186,695,440 / 7,264,414 EDGAR-filed shares = exactly $25.70, the reported previous_close, against a $26.10 last price. Recomputed to $189.6M in company.md C.4.
max_52_week_position fires the same previous-close bug: BPIQ's 0.51135 equals (25.70 - 4.98)/(45.50 - 4.98) to five decimals; the last-price computation gives 0.5212. Both known failure modes fire on the same ticker.
enterprise_value does not reconcile: $172,089,863 against market_cap $186,695,440 implies subtracting $14,605,577, matching neither the row's own cash field nor the current balance sheet. Recomputed.
cash, ttm_burn, qtr_burn and monthly_burn are all stamped 2026-03-31 - correct for that date ($16.379M matches the reported Q1 $16.4M) but three and a half months stale, predating the Ascendant licence payment, two financings and the entire change of management.
short_float 0.09563 is roughly 2x the FINRA-derived figure (349,876 shares at the 2026-07-31 settlement = 4.82% of shares outstanding, ~6.0% of estimated float) under either reading; no unit conversion reconciles them. FINRA used throughout.
The options chain was never adjusted for the 1-for-30 reverse split of 2026-04-23: every listed strike is at or below $5.00 against a $26.10 spot, so an at-the-money straddle is not merely noisy but impossible to construct. Three of the four expirations BPIQ returned (2026-05-15, 2026-06-18, 2026-07-17) had already passed at sweep time. Every call IV prints the 0.01488 floor artifact.
fetch_company_hedge_fund_holdings returned one block labelled 2026Q3 whose value/shares ratios ($31.54 and $31.55) agree with each other but match no quarter-end close - CUE traded near $31.5 in late July 2026. Following the CNTB precedent, share counts were used and the dollar values discarded.
fetch_company_insider_transactions did NOT exhibit the empty-result failure recorded on TLSA and CNTB: 250 rows came back, 2019-11-01 to 2026-07-30. It does lag the filing feed at the edge - EDGAR shows Form 4s on 2026-08-14 and 2026-08-17 that the connector does not carry - so the August flow was read from EDGAR and the press feed instead.
FINRA's 2026-07-15 settlement returned empty while 2026-06-30 and 2026-07-31 returned data - publication lag, not a missing ticker, exactly as 02-connectors.md describes.
data/prices/CUE.json does not exist. The price cache does not cover this ticker, so the 52-week position was hand-computed per 02-connectors.md rather than derived in lib/prices.mjs, the priced-in run-up driver fell back to its neutral default, and no run-up settlement can be resolved against committed price history until scripts/fetch-prices.mjs covers CUE.

Price cache

Derived from data/prices/CUE.json — not a sweep row, so it carries no CALLED/BLOCKED state

First bar	2021-08-19
Last bar	2026-08-19
Bars	1255
Fetched	2026-08-19

IMMP Cleared

Company tier

Swept 2026-08-19 · called once, reused by every program

BPIQ:fetch_company_info	Called
BPIQ:fetch_company_drugs	Called
BPIQ:fetch_company_historical_catalysts	Called
BPIQ:fetch_company_options_data	Called
BPIQ:fetch_company_insider_transactions	Called
BPIQ:fetch_company_press_releases	Called
BPIQ:fetch_company_hedge_fund_holdings	Called
BPIQ:fetch_company_earnings_transcript	Not called
EDGAR:submissions	Called
EDGAR:companyconcept dei:EntityCommonStockSharesOutstanding	Called
FINRA:consolidatedShortInterest	Called
Yahoo:options chain	Blocked

Program tier — efti-pembrolizumab-soft-tissue-sarcoma

BPIQ drug id 17452 · CLEARED

CT.gov:search_trials	Called
CT.gov:get_trial_details	Called
CT.gov:search_investigators	Called
PubMed:search_articles	Called
PubMed:get_article_metadata	Called
Open Targets:search_entities Rate limit exceeded for client: global	Blocked
ChEMBL:compound_search	Called
web_search:peak sales	Called
web_search:competitive	Called
web_search:exclusivity + royalty	Called
web_search:analyst	Called
EDGAR:full-text search	Not called
Europe PMC:search	Not called
CTIS:search	Not called

Open Targets BLOCKED on all three attempts (query sets ["LAG3","soft tissue sarcoma"], ["LAG3"], ["LAG3","eftilagimod alfa"]) - the standing platform throttle. Cost: no independent human-genetics validation of LAG-3, so the target-validation row rests on clinical precedent alone. Smaller than usual here because eftilagimod alfa is an MHC class II agonist rather than a LAG-3 blocker, so LAG-3 genetics would not validate what this drug does.

PubMed's null-authors bug did NOT fire: all ten articles returned complete author lists with affiliations, which is what made the kol tables possible at all.

No EFTISARC-NEO publication exists in PubMed fifteen months after the primary endpoint was announced - 10 of 10 hits on 'eftilagimod' and none is this trial. A real absence, not a connector failure.

The catalyst's own endpoint is unregistered: health-related quality of life appears among neither the primary, the seven secondary, nor any 'other' outcome measures on NCT06128863 (other_outcomes: null), and the instrument is not disclosed in the abstract-acceptance release. The settlement definition therefore names a threshold generically rather than against a pre-specified analysis.

CT.gov lists no investigators for NCT06128863: no overall officials, no principal investigator, single Warsaw location with contacts null. search_investigators against the sponsor institution returned 51 investigators across 18 OTHER sarcoma trials and none on this one. The two names in kol.investigators are inferred from institutional authorship and tagged UNVERIFIED.

No conflict-of-interest statements were retrievable - PubMed get_article_metadata returns affiliations but no COI field on either article used, so every conflicts_checked entry records where the search was made and what it could not establish.

ChEMBL compound_search succeeded but supplies no selectivity data (structure_type NONE, molecule_properties null, no SMILES) because efti is a recombinant fusion protein; its orphan flag reads 0 against an FDA Orphan Drug Designation granted 2026-04-15, i.e. stale.

The modelled readout estimate is uninformative and recorded anyway: primary completion 2025-04-30 plus the default two-to-four-month lag gives a window that closed in August 2025, because the method dates the trial's primary readout, which already happened. data/benchmarks/readout-lag.json is empty so the default applies.

A CONFIRMED readout-lag observation was produced and NOT written down (outside this lane): NCT06128863 primary completion 2025-04-30 against topline 2025-05-27 = 27 days, both from primary sources.

No price cache for this ticker: the priced_in run-up driver was scored on a synthetic two-bar stand-in carrying BPIQ's reported 52-week high and low rather than a committed series, and two rows of company.md C.7 cannot be sized at all.

BPIQ cash is denominated in AUD while market_cap is in USD and neither field says so: cash 99,129,499 at 2026-03-31 reconciles as A$99.13M against the A$68.87M the company reported at 2026-06-30 after a A$38.9M quarterly outflow; read as USD it reconciles with nothing. Mixing the two silently produces a negative enterprise value, which is what BPIQ reports.
enterprise_value -15,080,770 is unreconcilable against its own inputs, implying US$73.6M of cash netted - neither the June figure (US$48.8M) nor the March figure (US$70.2M) at AUD/USD 0.7083. Recomputed from the EDGAR-filed share count, spot and reported cash: US$9.6M.
has_catalyst is TRUE on row 13791, a Failed/Discontinued row with a null catalyst_date, five months after TACTI-004 was stopped. The CNTB stale-flag bug reproduced on a second ticker. Three of nine rows read as pending catalysts; two are.
Both of BPIQ's move fields for 2026-04-15 (FDA orphan designation) are unusable and contradict each other: open_price_gap_percent 2.20839 against intra_day_price_change_percent -0.46139 on a day the press feed describes as a surge.
BPIQ returns null for both move fields on the 2025-05-26 EFTISARC-NEO primary-endpoint row - the closest analogue to the analysed catalyst - and omits the 2025-10-20 ESMO 2025 presentation of the same data entirely. The historical-catalyst feed is not a complete event list for this ticker.
fetch_company_insider_transactions returned zero rows on two attempts and this is a LEGITIMATE EMPTY, not the TLSA/CNTB silent failure: the EDGAR submissions feed carries six Form 3 initial ownership statements in April 2026 and no Form 4 transaction reports at all in the visible window.
BPIQ short_float 0.00836 AGREES with FINRA/EDGAR (1,195,547 ADS at settlement 2026-07-31 over 146,038,958 ADS = 0.82%), unlike the CNTB case where the same comparison was off by 3.4x.
Options chain carries NO strike at or below spot on any expiry (lowest $0.50 against $0.40), so an at-the-money straddle is not constructible; 12 of 28 contracts sit on the IV ceiling artifact 9.99512 and 6 more on the floor artifact 0.01488. The nearest expiry to the catalyst (2026-10-16) expires three days BEFORE it.
No data/prices/IMMP.json exists, so 03a C.4's required lib/prices.mjs derivation could not be run; the read-only Yahoo chart cross-check returned the body 'Too Many Requests'. The 52-week range falls back to BPIQ's own high/low fields (max_52_week_position is not read at all).
Five distinct spellings of one molecule across nine drug rows, including two leading-space and three trailing-space variants. Names trimmed; nothing keyed on them.
BPIQ insider_own and tute_own carry no timestamp, so 'as of' for the ownership split is unknown rather than current.

Price cache

Derived from data/prices/IMMP.json — not a sweep row, so it carries no CALLED/BLOCKED state

First bar	2021-08-19
Last bar	2026-08-19
Bars	1255
Fetched	2026-08-19

CMPX Cleared

Company tier

Swept 2026-08-19 · called once, reused by every program

BPIQ:fetch_company_info	Called
BPIQ:fetch_company_drugs	Called
BPIQ:fetch_company_historical_catalysts	Called
BPIQ:fetch_company_options_data	Called
BPIQ:fetch_company_insider_transactions	Called
BPIQ:fetch_company_press_releases	Called
BPIQ:fetch_company_hedge_fund_holdings	Called
EDGAR:submissions	Called
EDGAR:companyconcept	Called
FINRA:consolidatedShortInterest	Called
Yahoo:options_chain	Not called

Program tier — tovecimig-biliary-tract-cancer

BPIQ drug id 17398 · CLEARED

CT.gov:search_trials	Called
CT.gov:get_trial_details	Called
CT.gov:search_investigators	Called
PubMed:search_articles	Called
PubMed:get_article_metadata	Called
PubMed:get_full_text_article	Called
OpenTargets:search_entities	Called
ChEMBL:compound_search	Called
web_search:peak_sales	Called
web_search:competitive	Called
web_search:exclusivity_royalty	Called
web_search:analyst	Called
EDGAR:full_text_search	Not called
EuropePMC:search	Not called
CTIS:search	Not called

CT.gov carries no investigator names for NCT05506943: search_investigators returned zero rows on two query shapes (condition + investigator_name 'Javle'; condition + institution 'Compass Therapeutics') and get_trial_details returned no overall-official or contact block. The kol investigator list is built instead from the COMPANION-002 design paper's author list, cross-checked against the trial's own CT.gov site list - a weaker chain than a registry read.

No conflict-of-interest statement exists for any named investigator: neither the PubMed metadata payload nor the PubMed Central full-text render of PMID 38861293 / PMC11509068 carries a financial-disclosure section. Every 'no competitor relationship found' entry rests on a search with nothing to search, which is recorded rather than treated as evidence of absence.

No independent voice on this endpoint could be found at all. The molecule's entire PubMed footprint is four papers - three the originator ABL Bio's own preclinical or pharmacokinetic work, one the sponsor's protocol description - and COMPANION-002's 34 sites enrol most of the US academic centres specialising in this disease, so its leading clinicians are largely investigators on it. kol.judgement.endpoint_supported is UNKNOWN in consequence.

No peer-reviewed publication of any tovecimig CLINICAL result exists. Every efficacy and safety figure in the analysis traces to a company press release or a congress abstract.

No composition-of-matter patent number, filing date or expiry could be located by the exclusivity web search. The IP assessment rests on the verified seven-year orphan-drug exclusivity alone.

The ABL Bio licence terms (tiered single-digit royalty, up to ~$399M of oncology milestones) come from consistent public summaries of the November 2018 agreement, not from the agreement or an SEC exhibit.

No published analyst peak-sales estimate for tovecimig in second-line biliary tract cancer was found, so B.3a is built bottom-up and nothing in it is a pass-through of a third-party figure.

Two independent published US biliary tract cancer incidence figures disagree (18,600 per the American Cancer Society via PMID 38861293; ~12,130 per SEER-based analyses) and are used as the two ends of a range rather than averaged.

ChEMBL returned no molecular properties and no bioactivity data. This is correct rather than a gap for an antibody, whose selectivity question ChEMBL's small-molecule tables answer poorly, and it costs nothing here.

The regulatory tier was not called, correctly: the catalyst is a congress data presentation and no marketing application has been submitted, so there is no regulatory docket to query.

finance_updated_at is 2026-06-30, so BPIQ's cash and burn fields are quarter-end figures rather than live ones.
market_cap fires the previous-close bug: $430,410,144 / 180,087,915 EDGAR-filed shares = $2.3900 exactly, against a $2.48 last price. Recomputed to $446.6M. The MNOV pattern (current share count, stale price), not the CNTB pattern (current price, stale share count).
max_52_week_position fires the same previous-close bug: 0.14801 = (2.39 - 1.61) / (6.88 - 1.61). Recomputed to 0.16509.
enterprise_value cannot be reconciled against the same payload's cash field: $245,757,083 implies subtracting $184,653,061 against a reported cash of $179,879,000, leaving $4,774,061 unexplained. Recomputed to $266.7M.
fetch_company_insider_transactions returned 113 rows and agrees with the EDGAR Form 4 feed on both recent clusters, so the empty-result bug recorded on TLSA and CNTB did NOT fire on this ticker.
Hedge-fund block labels are FILING quarters on this ticker, verified by implied price rather than assumed: the 2026Q3 block returns ~$2.19/share across six rows, 2026Q2 returns $5.29 across ten, 2026Q1 returns $5.37. Only the filing-quarter reading fits, because $2.19 is post-collapse and the other two are pre-collapse. Each block's holdings are one quarter older than its label.
short_float (0.30106) disagrees with FINRA and was not used: the FINRA 2026-07-15 settlement gives 21.8% of shares outstanding or 24.9% of estimated non-affiliate float, neither of which is 30.1%. The disagreement could not be resolved because the 2026-07-31 settlement request timed out twice.
The options chain carries NO strike at or near the $2.48 spot (strikes jump 2.00 to 3.00), so an at-the-money straddle cannot be constructed at all - the same state observed on CNTB and MNOV on 2026-08-13, now on a third ticker. Both implied-volatility artifacts are present on the near expiry.
data/prices/CMPX.json does not exist. Not a coverage row (the price cache never lands here), recorded as a note because it forced the 03a C.4 by-hand fallback for the 52-week range and left one run-up driver's input substituted.

Price cache

Derived from data/prices/CMPX.json — not a sweep row, so it carries no CALLED/BLOCKED state

First bar	2021-08-19
Last bar	2026-08-19
Bars	1255
Fetched	2026-08-19

RAPP Cleared

Company tier

Swept 2026-08-19 · called once, reused by every program

BPIQ:fetch_company_info	Called
BPIQ:fetch_company_drugs	Called
BPIQ:fetch_company_historical_catalysts	Called
BPIQ:fetch_company_options_data	Called
BPIQ:fetch_company_insider_transactions	Called
BPIQ:fetch_company_press_releases	Called
BPIQ:fetch_company_hedge_fund_holdings	Blocked
BPIQ:fetch_company_earnings_transcript	Not called
EDGAR:submissions	Called
EDGAR:companyconcept	Called
FINRA:consolidatedShortInterest	Called
Yahoo:options	Not called

Program tier — rap-219-bipolar-mania

BPIQ drug id 18871 · CLEARED

CT.gov:search_trials	Called
CT.gov:get_trial_details	Called
PubMed:search_articles	Called
PubMed:get_article_metadata	Called
OpenTargets:search_entities Rate limit exceeded for client: global	Blocked
ChEMBL:compound_search	Called
web_search:peak_sales	Called
web_search:competitive	Called
web_search:exclusivity_royalty	Called
web_search:analyst	Called
EDGAR:full_text_search	Not called
EuropePMC:search	Not called
CTIS:search	Not called
CT.gov:search_investigators	Called
PubMed:conflict_search_on_investigator	Called

OPEN TARGETS BLOCKED, four attempts, identical global rate limit each time. Costs the independent human genetic evidence on whether CACNG8 or AMPA-receptor genes are implicated in bipolar disorder. Target validation is therefore scored on the sponsor's own preclinical rationale plus one clinical result in a different disease (epilepsy), and tagged UNVERIFIED.

ChEMBL returned a LEGITIMATE EMPTY RESULT for RAP-219 (count 0) - a CALLED state, not a block, and ordinary for an unapproved small molecule whose structure is unpublished. But it means the regional-selectivity claim the entire mechanism rests on has no independent bioactivity data behind it. More expensive here than the tool's usual caveat suggests, because unlike the antibody programs elsewhere in this corpus this is a small molecule whose whole thesis IS selectivity.

THE DOSE AND REGIMEN ARE NOT DISCLOSED by any source. The registry says only 'RAP-219 tablets administered orally, once daily for 21 days'. Against a ~22-day half-life and a 21-day primary endpoint this is the single most consequential missing fact in the analysis, and no public source resolves it.

CT.gov search_investigators returned zero investigators, so the kol block names ONE person - the trial's overall official, recovered from the registry record itself - where twenty sites are listed. The PubMed conflict search on that name returned only pre-1990 articles by a different author of the same surname and initials, so conflicts_checked records a search made, never a certification that no conflict exists.

NO PEER-REVIEWED PRIMARY PUBLICATION OF ANY RAP-219 TRIAL EXISTS. PubMed returned three hits in total; two are Eilat conference-summary reviews in Epilepsia about EPILEPSY, one is unrelated 2017 screening work. Every efficacy figure in the analysis comes from a company press release, including the 77.8% seizure reduction and 24% seizure-free rate that carry the target-validation argument. The PubMed authors-null bug did NOT fire here - all authors returned correctly.

The YMRS competitive benchmark figures (asenapine -10.8 / olanzapine -12.6 vs placebo -5.5 at day 21; response rates 42.5% / 52.4% / 29.4%) are WEB ESTIMATE from a search over the published 3-week trial literature, not read from the primary papers. They set the bar the readout is judged against, so the limit is stated: direction and rough magnitude are well established across many trials, but no individual figure was confirmed against its own source document.

NO PATENT OR EXCLUSIVITY INFORMATION WAS LOCATED. The dedicated exclusivity search returned the Tenacia Greater China licence terms in full and nothing on composition-of-matter claims or expiry dates. The IP row and the market-exclusivity row both record UNKNOWN rather than a guess.

The original enrolment target for NCT07046494 is not disclosed anywhere read this sweep, against a company statement that target enrolment was INCREASED mid-trial to 253 actual. The size of the increase - and therefore how much statistical power was added - cannot be established.

No committed price cache for this ticker, so the run-up exit is null, the priced_in driver reads a hand-computed 52-week position rather than one derived from daily bars, and the spot comes from BPIQ rather than from a cache.

First analysis of this ticker. No committed price cache exists - data/prices/RAPP.json is absent - so the 52-week range and position in C.4 are recomputed by hand from BPIQ's own fields under 03a C.4's stated fallback rather than derived in node with lib/prices.mjs. The price cache is not a sweep row and never lands here; recorded as a note because it changes how one company-tier figure was produced and how the program's runup priced_in driver was scored.
market_cap fires the documented previous-close bug in its stale-price form: $2,351,415,808 / 47,978,286 EDGAR-filed shares = exactly $49.0102, the previous_close, against a $48.94 last price. Fourth ticker on which this has been reproduced.
max_52_week_position fires the same bug: 0.98624 is exactly (49.01-13.90)/(49.50-13.90). Last-price computation gives 0.98427. At this position in the range the practical effect is small, which is why it would be easy to pass through unnoticed elsewhere.
enterprise_value cannot be reconciled against the record's own cash field: $1,897,136,608 implies subtracting $454.28M against a cash field of $436,077,000, an $18.2M residual matching no balance-sheet item read this sweep.
tute_own reads 1.08713 - 108.7% institutional ownership, impossible, and with the 4.47% insider figure it exceeds 113%. Discarded under 01-rules.md rule 6 rather than converted.
short_float reads 0.06960 against a FINRA-derived 4.69% of shares outstanding (4.91% of estimated float) - 1.48x the larger reading, not a factor-of-100 slip, and no reading reconciles them.
fetch_company_insider_transactions returned ZERO ROWS across two attempts while 85 Form 4 filings sit in the EDGAR feed, 35 of them in 2026 and the latest dated 2026-08-12. THIRD TICKER on which this documented bug has been reproduced (after TLSA and CNTB). The call completes rather than erroring, so the state is CALLED with an empty result; the flow was read from EDGAR and the press feed instead.
ttm_burn ($77,218,000) is inconsistent with qtr_burn ($41,455,000) - four quarters at the current rate would be $165.8M. Explained rather than buggy: Q1 2026 carried a $20M upfront licensing payment and the Phase 3 programme only began spending in Q2. A naive runway off ttm_burn would be wrong by more than a factor of two.
Drug names are dirty as documented: four spellings of one molecule across five of six rows, and TWO ROWS SHARE THE IDENTICAL drug_name 'RAP-219' differing only in indication (18870 focal epilepsy, 18871 bipolar mania). The integer id is the only safe key on this ticker.
The pipeline's has_catalyst flags do not locate this company's value: the two Phase 3 epilepsy trials carrying most of the enterprise value (NCT07563881 n=333, NCT07594119 n=312, primary completion 2029-07) have no BPIQ row with a data catalyst at all.
The regulatory tier was correctly NOT CALLED - the only catalyst analysed under this ticker is a clinical-trial readout, not a regulatory decision.

Price cache

Derived from data/prices/RAPP.json — not a sweep row, so it carries no CALLED/BLOCKED state

First bar	2024-06-07
Last bar	2026-08-19
Bars	551
Fetched	2026-08-19

Key learning

'No approval received' is a transient error, not a permission gate; retry before recording BLOCKED. Open Targets' shared 'client: global' rate limit is intermittent rather than permanent: on 2026-08-12 it blocked the IVA program sweep across three attempts while the OLMA sweep called search_entities successfully the same day — retry, and record BLOCKED only when the attempts are exhausted. BPIQ tools can also throw transient 502s that clear on retry (TLSA options), and can return legitimate empty results that still count as CALLED (TLSA insider/hedge-fund/earnings).