RAPP — Rapport Therapeutics, Inc.

Company context · prepared 2026-08-19 · company sweep 2026-08-19 · USD · framework v5.10.3

How to read this. Written for a reader who has not studied pharmacology or finance. General terms are defined once in framework/04-glossary.md. Tags: [VERIFIED — source] / [UNVERIFIED] / [WEB ESTIMATE — source, date].

Program analyses for this company are the sub-folders listed in C.2.

C.0 Company-tier coverage — CLEARED

Every mandatory company-tier row was called. Two rows are recorded honestly as failures: BPIQ’s insider-transaction connector returned nothing while 85 Form 4 filings sit in the EDGAR feed, and the optional hedge-fund row was refused twice by the proxy. The regulatory tier was not called at all, which is correct rather than a gap — 02-connectors.md makes that tier conditional on a regulatory catalyst (a decision by the Food and Drug Administration, an advisory committee, or a similar event), and the only catalyst analysed under this ticker is a clinical-trial readout.

There is no committed price cache for this ticker: data/prices/RAPP.json does not exist. 02-connectors.md puts the cache outside any analysis run, on its own cadence, so this sweep did not create one. The consequence is confined to C.4 and is named there.

C.1 What the company is

Rapport Therapeutics is a clinical-stage biotechnology company in Boston, Massachusetts, developing small-molecule medicines for disorders of the brain and nervous system [VERIFIED — BPIQ fetch_company_info, 2026-08-19]. It has no approved product and no product revenue. Its only revenue to date is a $20 million upfront payment from a regional licensing deal booked in the first quarter of 2026 [VERIFIED — company press release, 2026-03-09, and the Q1 2026 results reported 2026-05-07].

It is effectively a single-asset company. Five of the six programs BPIQ lists are the same molecule, RAP-219, in different diseases and different formulations; the sixth, RAP-641, has not yet filed to begin human testing. The share price this document records is therefore a price on one molecule.

The lead use is drug-resistant focal epilepsy, where two large confirmatory trials began in 2026 and read out in 2029. The event this repository is analysing is a secondary use: acute mania in bipolar I disorder, a proof-of-concept study reading out in October 2026. The program document is at rap-219-bipolar-mania/.

Spot for every figure below: $48.94, read 2026-08-19 [VERIFIED — BPIQ fetch_company_info, 2026-08-19].

C.2 Pipeline — every program

Every row returned by fetch_company_drugs on 2026-08-19 is listed, including the two with no catalyst flag and the one whose application to begin human testing was placed on hold. Catalysts are reported as catalyst_date_text; the “to a day?” column is yes only where that text names a day, per 01-rules.md rule 23. None of them does.

One thing this table does not show, and it matters. The company’s largest value driver — the two confirmatory Phase 3 epilepsy trials, FOCUS 1 (NCT07563881, 333 patients) and FOCUS 2 (NCT07594119, 312 patients), both begun in 2026 with primary completion in July 2029 — has no row of its own with a data catalyst [VERIFIED — ClinicalTrials.gov search_trials on intervention RAP-219, 2026-08-19]. Row 19646 is a different indication, and row 18870 is the earlier Phase 2 and its extension. A reader who counts materiality only across the rows above will conclude that the bipolar readout is a larger share of this company than it is. It is not: most of the $1.91 billion enterprise value in C.4 is a bet on epilepsy trials that report in 2029.

C.3 Financial position

Dilution history. Every raise in the last 24 months, plus the live facilities that have not yet been drawn.

No withdrawn offering was found in the filing feed or the press feed.

Is a raise likely before the October 2026 catalyst? Almost certainly not before it; quite possibly within days of it. The balance sheet does not need money — $436.1M against a burn of roughly $13.6M a month. But this company’s demonstrated pattern is to raise on good news, not before it: the 2025-09-08 Phase 2a topline release and the proposed public offering were announced the same day, and the offering priced 48 hours later [VERIFIED — press feed items dated 2025-09-08 and 2025-09-10; 424B5 filed 2025-09-10]. An unlimited automatic shelf and an undrawn $110M at-the-market facility are both live. A reader positioned for a positive readout should expect the supply that follows it.

C.4 Valuation frame

Why the 52-week position was computed by hand rather than from the price cache. 03a’s standing instruction is to derive this row in node from data/prices/<TICKER>.json with lib/prices.mjs. That file does not exist for RAPP — this is the ticker’s first analysis, and the price connector runs on its own cadence outside an analysis run (02-connectors.md). 03a provides for exactly this case: recompute the formula by hand for a ticker the cache does not cover. That is what the two rows above do, from BPIQ’s own fifty_two_week_high and fifty_two_week_low.

The consequence is worth stating plainly rather than hiding. The hand computation reproduces 01-rules.md rule 11’s formula but inherits whatever convention BPIQ’s own high and low fields use, and it cannot be recomputed for a past date the way a cache can. It is also the input the run-up priority score’s priced_in driver reads in the program document, so that driver’s basis names the same limitation. BPIQ’s own max_52_week_position field reads 0.98624 and is not used: that value is exactly $(49.01 − 13.90) ÷ (49.50 − 13.90)$, the previous close, so the documented previous-close bug fires here too and the field is discarded rather than corrected.

The plain reading: this share price is at the top of its own one-year range and at an all-time high. Three press items independently record new all-time highs on 2026-06-30 ($42.36), 2026-08-06 ($44.38) and 2026-08-12 (a new one-year high) [VERIFIED — BPIQ press feed], and the $49.50 high is above all of them, so the stock has continued making highs through the sweep week. Whatever else is true of this event, very little of the downside is already discounted.

C.5 Ownership and flow

Short-interest trend. It spiked into mid-July and then unwound: 2,246,097 shares at the 2026-06-15 settlement, 2,260,158 at 2026-06-30, 2,690,734 at 2026-07-15 (+19.05%), and 2,248,235 at 2026-07-31 (−16.45%) [VERIFIED — FINRA consolidatedShortInterest, three settlements read 2026-08-19]. Days to cover rose from 4.62 to 7.51 over the same span, because average daily volume fell from 488,981 shares to 299,167 while the position was roughly flat. Read plainly: a short position was put on and taken off again during July, and what remains is small. At 4.7% of shares outstanding this is not a crowded short, and there is not much fuel here for a squeeze.

Insider flow, read from filings because the connector returned nothing. BPIQ’s fetch_company_insider_transactions returned zero rows across two attempts. The EDGAR feed carries 85 Form 4 filings, 35 of them since 2026-01-01, the most recent dated 2026-08-12 [VERIFIED — EDGAR submissions, CIK 0002012593, read 2026-08-19]. The press feed names the transactions behind them.

The pattern is uniform and it is one-directional:

• Chief executive Abraham Ceesay sold on 2026-01-23 ($285,000), 2026-02-20 (5,833 shares), 2026-07-01 (8,220 shares), 2026-07-15 ($328,436) and 2026-08-12 (28,958 shares and a further 22,541 shares) — the last of these in the sweep week, at or near the all-time high.
• Chief operating officer Cheryl Gault sold 10,000 shares on 2026-04-01 and 10,000 more on 2026-07-02.
• Chief development officer Krishnaswamy Yeleswaram sold on 2026-01-05 (2,840 shares), 2026-03-26 ($303,355), 2026-04-09 (a small block) and 2026-04-21 ($808,844).
• Director David Bredt sold 8,500 shares on 2025-12-17. A director received a grant of 15,200 options on 2026-06-11.

Every one of these is described in the feed as executed under a Rule 10b5-1 plan — a schedule set in advance, which insulates the seller from an allegation of trading on inside information and which therefore says much less about conviction than a discretionary sale would. [VERIFIED — BPIQ press feed items dated as listed, 2026-08-19]

What is absent is the finding. Across 227 press items and 85 Form 4 filings, this sweep found no open-market purchase by any officer or director. Uniform pre-planned selling with no buying is the neutral case, not the bearish one; but it is emphatically not the conviction-buying pattern that sometimes precedes a positive readout, and a reader should not supply one from imagination.

A ten-percent holder is distributing. Third Rock Ventures, the venture firm that founded the company, sold repeatedly through April and May 2026: 426,005 shares reported 2026-04-23, ~$17.1M reported 2026-04-22, ~$11.09M reported 2026-05-11, ~$5.38M reported 2026-05-14, and 836,136 shares in total across April–May 2026 [VERIFIED — BPIQ press feed, items dated as listed]. These too are described as Rule 10b5-1 sales. This is systematic distribution by the largest early holder into a rising price.

Mechanical buying on the other side. The shares were added to the Russell 2000 Growth, Russell 3000E Growth and Russell Microcap Growth benchmark indices on 2026-06-29 [VERIFIED — BPIQ press feed, three items dated 2026-06-29]. Index inclusion creates buying that has nothing to do with the readout, and it coincides with the June-to-August advance.

Fund positioning. Six funds hold the name [VERIFIED — BPIQ fetch_company_info hf_holding_count, 2026-08-19]. Nothing further: the holdings connector was refused twice (C.0), so which funds, how much, and whether they added or trimmed are all [UNVERIFIED] this sweep. Two individual positions surface in the press feed — Baker Bros. Advisors at ~$45.95M (reported 2026-03-15), Capital International Investors buying 800,000 shares (reported 2026-03-12) [WEB ESTIMATE — MarketBeat, 2026-03] — and are recorded as third-party reports, not as a positioning read.

C.6 Options chain and its readability

Plain takeaway: the chain says the market expects a very large move, and it is not liquid enough to say how large with any precision. At the November expiry that spans the readout window, the cost of a position that profits from a move in either direction runs somewhere between a third and half of the share price — a bracket of roughly 33% to 51%, not a number. That bracket is so wide because the bid and the ask on the two contracts involved are $8.00 and $12.50. Open interest at the strike nearest the share price is ten contracts on one side and none on the other. Anyone quoting a single “implied move” figure off this chain is quoting the midpoint of a spread nobody has traded.

One line in the chain is worth reading on its own, and it is not the at-the-money one. The November $40 put — protection struck 18% below the share price, expiring three weeks after the readout window opens — carries 1,500 contracts of open interest, by far the largest single line at any strike in any expiry. Somebody has bought a meaningful quantity of downside protection across this event.

Artifact check. The documented implied-volatility floor of ~0.01488 is present on six strikes of the 2026-08-21 expiry ($25, $32.50, $35, $37.50, $40 and $42.50), all deep in the money with two days to expiry, and every one of them also reports a delta of exactly 1 and a gamma and vega of exactly 0. All six are treated as null. The ceiling artifact (~9.99512) does not appear, though the 2026-08-21 $15 call reads 7.20 — high, but below the documented ceiling and explicable at two days to expiry. None of the discarded strikes is one this document reads.

C.7 Reaction to past catalysts

Five rows returned. Two of them share a date and carry null price fields, so three are usable. Each date was cross-checked against the press feed.

The range for a clean, positive, data-driven move on this ticker is not establishable from three rows, and the largest of them is not a clean analogue. The +39.6% of 2025-09-08 was the lead indication’s proof of concept, at a share count 25% smaller and a share price around $9 before the gap, in a company that had no other value driver at the time. The bipolar readout is a secondary indication at a $2.35 billion market capitalisation with a funded Phase 3 program behind it. The program document’s scenario ranges use this row as one anchor and say plainly why it is an upper bound rather than a forecast.

This table contains no negative surprise at all. Not one row is a miss. The nearest observations of how this equity trades bad news come from outside the catalyst table: −7.2% on 2026-01-02, −6.9% on 2026-01-13 and −7.4% on 2026-01-18 around the accelerated Phase 3 announcement and the at-the-market filing, and −7.7% / −7.9% on 2026-03-13 and 2026-03-14 on the China licensing deal and the automatic shelf [VERIFIED — BPIQ press feed, items dated as listed]. Those cluster at −7% to −8% and are all dilution or expectation-reset events, not clinical failures. A clinical miss should be expected to be worse than any of them.

C.8 Company data-quality flags

1. market_cap is computed off the previous close. BPIQ reports $2,351,415,808. Divided by the EDGAR-filed share count of 47,978,286 that is exactly $49.0102 — the previous_close field of the same record — against a last_price of $48.94. This is the documented bug in 02-connectors.md § Field-level bugs, now reproduced on a fourth ticker, and in its stale-price rather than its stale-share-count form. C.4 multiplies the filed share count by the price it actually wants instead.
2. enterprise_value cannot be reconciled against the record’s own cash field. BPIQ reports $1,897,136,608, which against its own market capitalisation implies subtracting $454.28M. Its own cash field is $436,077,000, and that figure is itself exactly the sum of the two EDGAR-filed balance-sheet items. The $18.2M residual matches no item read this sweep — not the operating lease liability ($11.484M), not total liabilities ($36.056M), and not any prior quarter-end cash balance ($476.782M at 2026-03-31, $490.539M at 2025-12-31). C.4 recomputes rather than smoothing the gap over.
3. short_float disagrees with FINRA by a factor no unit conversion explains. BPIQ reads 0.06960. The FINRA count of 2,248,235 shares at the 2026-07-31 settlement is 4.69% of shares outstanding or 4.91% of estimated non-affiliate float. BPIQ’s figure is 1.48× the larger reading, so it is not a percentage-versus-fraction slip that a factor of 100 would fix, and there is no reading under which the two agree. C.5 uses the FINRA figure, whose units are known.
4. max_52_week_position uses the previous close. BPIQ reads 0.98624, which is exactly $(49.01 − 13.90) ÷ (49.50 − 13.90)$. The last-price computation gives 0.98427. The field is not read; C.4 recomputes. Note that at this position in the range the bug’s practical effect is small — both readings say the same thing — which is precisely why it would be easy to pass through unnoticed on a ticker where it is not.
5. tute_own reads 1.08713 and is discarded. As a fraction that is 108.7% institutional ownership. Combined with the 4.47% insider figure it exceeds 113%. 01-rules.md rule 6 requires rejecting a broken third-party figure rather than passing it through, so C.5 records the row as not usable and derives nothing from it.
6. 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 reproduction of the documented connector bug. The call completes rather than erroring, so the coverage state is CALLED with an empty result — the emptiness is not evidence of no insider activity. C.5 reads the flow from EDGAR and the press feed instead.
7. ttm_burn is internally inconsistent with qtr_burn. BPIQ reports a trailing-twelve-month burn of $77,218,000 against a quarterly burn of $41,455,000 — four quarters at the current rate would be $165.8M. The explanation is real rather than a bug: the first quarter of 2026 carried a $20M upfront licensing payment and the Phase 3 program only began spending in the second quarter, so the trailing figure averages over a much lighter period. It is recorded here because a naive runway computed off ttm_burn would give 68 months and be wrong by more than a factor of two.
8. A syndicated press item dated 2026-04-21 reports a $250M offering at $26.00 that did not happen in April 2026. It is a re-publication of the September 2025 follow-on. No 424B5 was filed between 2026-01-07 and the sweep date. C.7’s 2026-04-21 row corrects this rather than treating it as a same-day bundle.
9. The pipeline’s has_catalyst flags do not locate this company’s value. The two Phase 3 epilepsy trials that carry most of the enterprise value have no BPIQ row with a data catalyst at all. This is not a bug in the feed — those readouts are three years away — but any materiality judgement drawn only from the rows in C.2 will overstate every one of them. C.2’s closing note says so directly.
10. No committed price cache exists for this ticker. data/prices/RAPP.json is absent, so C.4’s 52-week range and position are computed by hand from BPIQ’s fields under 03a’s stated fallback, rather than derived in node from the cache. The run-up priority score’s priced_in driver in the program document reads that same hand computation and its basis names the substitution.