VALN / s4v2-shigellosis — Shigella4V2 (S4V2) for the prevention of shigellosis

Program analysis · bpiq_drug_id 18578 · prepared 2026-08 · USD · framework v5.6.1 · Coverage: CLEARED

Company context, financials, ownership, options and the full pipeline: ../company.md.

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

Glossary

Executive summary

• What it is (one sentence): S4V2 is an injected vaccine that carries the sugar coats of the four most common Shigella types, each stitched onto a carrier protein inside a living E. coli cell, intended to prevent shigellosis — a bacterial diarrhoea that kills roughly 600,000 people a year and for which no vaccine has ever been licensed anywhere.
• The event and when (as disclosed): The first Phase 2 results, guided as “Mid-2026” and not to any day or month. The pivotal piece is a controlled human infection study (NCT06615375) in which vaccinated and placebo volunteers were deliberately infected with live Shigella sonnei. That trial’s registry primary-completion date, 2026-04-14, has already passed, and nothing had been announced as of 2026-08-11. This analysis places the readout window at 2026-08-13 to 2026-12-31, opening on the company’s scheduled half-year results date.
• The main reason it could work: Bioconjugation reliably produces functional, bacteria-killing antibodies against each Shigella sugar coat, and in the one previous challenge study on this platform the level of that antibody was statistically associated with protection. This trial additionally selects its dose in a first step before committing to the efficacy comparison in the second, which the earlier study did not do.
• The main risk: The direct precedent failed. Flexyn2a — the same platform, the same carrier protein, the same challenge model, run by the same academic network — reduced shigellosis by 30.2% and missed statistical significance (p=0.11). S4V2 must roughly double that effect on a study of similar size, against the same permissive endpoint definition that counts every moderate case.
• What it means for the stock: Little, in either direction. Valneva is a revenue-generating vaccine company whose share price is set by its Lyme disease program, not by this one. The shares trade at 7.3% of their 52-week range after falling 37% in a single session on the Lyme Phase 3 result, so a Shigella disappointment removes an option the market is barely paying for, and a Shigella success is capped by an unfunded Phase 3 and a contractual warrant strike. The expected value is $5.36 against a $5.30 spot, and its sign flips inside the probability band.

0. Program-tier coverage — CLEARED

Company-tier coverage is in ../company.md C.0.

Verdict: CLEARED. Every mandatory program-tier row was called. One row is BLOCKED — Open Targets, on the standing global rate limit — and the claim it would have supported is named above and is not one this analysis rests on. No row reads NOT CALLED.

A. Scientific & clinical assessment

A.1 Mechanism of action and scientific rationale

What the drug is. S4V2 is an injected vaccine, given as two shots into the muscle, mixed with an aluminium adjuvant called Alhydrogel [VERIFIED — ClinicalTrials.gov NCT06615375 and NCT06523231, read 2026-08-11]. Its active ingredients are the outer sugar coats — the O-antigens — of four kinds of Shigella bacteria: S. sonnei, and S. flexneri types 2a, 3a and 6 [VERIFIED — LimmaTech/Valneva releases; ClinicalTrials.gov NCT06615375 immunogenicity endpoints, which measure IgG against exactly those four].

How it works. A bare bacterial sugar is a poor vaccine. The immune system can make antibody against it, but the response is weak, fades quickly and barely works in infants, because sugars do not recruit helper T-cells. The standard fix, used for decades in vaccines against meningitis and pneumococcus, is to chemically staple the sugar to a protein. S4V2’s platform does the stapling a different way: an engineered E. coli cell is given the genes to build the Shigella sugar chain, the genes for a carrier protein (a detoxified Pseudomonas aeruginosa exotoxin A, “EPA”), and the gene for an enzyme (PglB) that joins the two together inside the living cell [VERIFIED — Ravenscroft et al., *Glycobiology* 2019, [DOI 10.1093/glycob/cwz044](https://doi.org/10.1093/glycob/cwz044), via PubMed]. The bacterium manufactures the finished vaccine molecule itself. The claimed advantages over chemical conjugation are a single production step, a uniform single-molecule product that is easier to characterise, and preservation of the sugar’s natural shape.

Once injected, the carrier protein recruits T-cell help, and the immune system makes serum antibody — specifically IgG — against each of the four sugar coats. That antibody is meant to reach the gut wall and bind Shigella on contact, where it kills the bacteria in two ways: by triggering the complement system, a cascade of blood proteins that punches holes in bacterial membranes, and by tagging the bacteria to be eaten by immune cells.

How well the target is validated. Better than the field’s track record suggests, and not fully. Three things are established. First, anti-LPS antibody is the right thing to raise: in the Flexyn2a challenge study, the level of S. flexneri 2a LPS-specific serum IgG was statistically associated with protection from disease (p=0.0016) and with a lower disease-severity score (p=0.002) [VERIFIED — Talaat et al., *EBioMedicine* 2021, [DOI 10.1016/j.ebiom.2021.103310](https://doi.org/10.1016/j.ebiom.2021.103310)]. Second, the antibodies the platform raises are functional, not merely present: rabbit serum raised against the quadrivalent bioconjugate bound and killed 22 of 41 vaccine-targeted Shigella isolates collected in Kenya, and cross-reacted with three serotypes the vaccine does not target [VERIFIED — Odundo et al., *mSphere* 2022, [DOI 10.1128/msphere.01020-21](https://doi.org/10.1128/msphere.01020-21)]. Third, the platform induces gut-homing antibody-secreting cells, meaning the response reaches the tissue where infection happens [VERIFIED — Clarkson et al., *EBioMedicine* 2021, [DOI 10.1016/j.ebiom.2021.103308](https://doi.org/10.1016/j.ebiom.2021.103308)].

What is not established is the step that matters commercially: that raising this antibody prevents enough disease to clear a statistical bar. The one time it was tested, it did not — see A.2.

The exact scientific step the next readout must prove. That two doses of S4V2 produce enough functional anti-S. sonnei antibody to significantly reduce the number of volunteers who develop shigellosis after being deliberately swallowed 1,500 live S. sonnei 53G bacteria, compared with volunteers who received placebo [VERIFIED — NCT06615375 primary outcome].

The honest scientific risk. Three specific things, in order of weight.

1. The predecessor failed on this exact question. Flexyn2a cut shigellosis by 30.2% and the result was not statistically significant. Everything about S4V2’s mechanism that could be called validated was equally true of Flexyn2a.
2. A tetravalent vaccine is being tested against one serotype. The challenge organism is S. sonnei only. Whatever S4V2 does against S. flexneri 2a, 3a and 6 is untested by this trial and will remain untested until a field study. A positive result proves one quarter of the product.
3. Spreading antigen across four components may weaken each one. A tetravalent vaccine at a given total dose delivers less of each sugar than a monovalent at the same total. Whether S4V2’s selected dose delivers as much S. sonnei O-antigen as Flexyn2a delivered of S. flexneri 2a is not disclosed [UNVERIFIED — the dose levels in NCT06615375 are described only as "high dose" and "low dose"; no microgram figures are public].

A.2 Clinical development plan, timeline, feasibility, resourcing

Trials

What the primary endpoint actually counts. A challenged participant is scored as having shigellosis if they have severe diarrhoea; or moderate diarrhoea plus either fever, or at least one at-least-moderate constitutional or gut symptom, or two or more vomiting episodes in 24 hours; or dysentery plus any of those same three [VERIFIED — NCT06615375 primary outcome description]. This is a broad definition that catches moderate illness, not only severe illness. It is the same broad definition Flexyn2a missed on while winning on the narrower severe-disease measures, and that is the single most important design fact in this document.

Feasibility matrix

Resourcing sufficiency. For this readout, entirely sufficient, and the reason is structural rather than financial: LimmaTech sponsors and pays for both Phase 2 trials, and Valneva assumes development only afterwards [VERIFIED — Valneva/LimmaTech partnership release, 2024-08-01]. The infant study additionally carries Gates Foundation support. Valneva’s cost to reach this decision point is therefore close to nothing beyond the €10 million upfront it has already paid.

The resourcing question that matters is the next one. If the results are positive, Valneva takes on all further development, chemistry-manufacturing-and-controls work, regulatory activity and worldwide commercialisation. A Shigella Phase 3 is a field efficacy trial in infants in low- and middle-income countries — thousands of participants over years. Valneva does not disclose a budget for it, has just cut 10–15% of its workforce and 25–35% of its operating expenses, and holds about $164M of cash against $215M of debt (see ../company.md C.3). A positive result creates a funding problem rather than solving one, and the realistic route is a partner, a public-health funder such as Gavi, CEPI or the Gates Foundation, or both.

A.3 Target label, target product profile, strategic questions, designations

A.3a Target label. Active immunisation for the prevention of shigellosis caused by Shigella sonnei and Shigella flexneri serotypes 2a, 3a and 6 — with two distinct populations in view: infants in endemic low- and middle-income countries, and adult travellers and military personnel from high-income countries.

A.3b Target product profile

A.3c Strategic Go/No-Go questions. The relevant set is Pre-Phase-III: both Phase 2 trials are run and the decision in front of the company is whether to take S4V2 into registrational development.

A.3d Regulatory designations.

• FDA Fast Track designation, granted 2024-10-16 to Shigella4V for the prevention of shigellosis [VERIFIED — Valneva/LimmaTech release, 2024-10-16, via the company press feed]. What it grants: more frequent written and in-person contact with the FDA during development, eligibility to submit a marketing application in pieces as they are completed (rolling review) rather than all at once, and eligibility to be considered for priority review later. It grants no evidentiary relief — the data still have to be there.
• No breakthrough therapy designation, no orphan designation, and no WHO prequalification pathway entry has been announced [VERIFIED — absence checked against the 265-item company press feed and the four web searches, 2026-08-11].

A.4 Target product profile valuation matrix

The calibration examples the framework offers here — roughly $150–300M of peak-sales potential per incremental month of overall survival in oncology, and 15–25% price premiums for oral over injectable formulations — are oncology and small-molecule reference points. Neither transfers to a prophylactic vaccine sold partly at tiered public-health prices, and neither is used as an input below.

A.5 Evidence quality and endpoints

Endpoints

A.5b Key opinion leaders.

Panel as of. 2026-08-11.

Investigators

No investigators are recorded, and that is a finding rather than an omission. ClinicalTrials.gov discloses no overall official and no site contact on either S4V2 trial — every contact field on all three US challenge-study locations and on the single Kenyan infant-study site is null. A separate condition-level search_investigators call across 50 shigellosis trials returned 26 named investigators, and not one of them is attached to NCT06615375 or NCT06523231; they belong to other sponsors’ trials. There is therefore no person this analysis can name as an investigator on this program’s pivotal trial without inventing the attachment, which rule 40 forbids.

Independent voices

Also empty, and for a reason worth stating plainly: in this field there are no independent voices. The complete indexed literature on this platform is six papers, and all six carry LimmaTech authors. The academic names that recur — Kawsar R. Talaat and David Sack at Johns Hopkins, Robert W. Kaminski and Kristen A. Clarkson at the Walter Reed Army Institute of Research, Mark S. Riddle and Chad K. Porter at the Naval Medical Research Center, Andrew J. Pollard at Oxford, Neil Ravenscroft at Cape Town — are every one of them co-authors with LimmaTech employees on the sponsor’s own studies, so none of them can be recorded here as independent. One of them carries a competitor relationship on top: Kawsar R. Talaat is the first author of the Flexyn2a challenge paper and is separately listed on ClinicalTrials.gov as the contact for EveliQure’s competing ShigETEC Phase 2 challenge study (NCT07049159) at the same Johns Hopkins Center for Immunization Research that is a listed S4V2 site. That is a real, checkable, dual relationship, and it is recorded here in prose because she is not an investigator on this program’s own trial and so has no row in the table above. This is exactly the case rule 40 anticipates — “in a small indication there commonly are none” — and the honest answer is an empty table with the search behind it on record, not a name promoted past what the evidence supports.

Judgement

Dissent

Empty, correctly: “UNKNOWN” is not a claim anyone can dissent from.

B. Commercial assessment

B.0 Current treatment algorithm

There is no treatment algorithm to slot into, because there is no preventive product at all. What happens to a person at risk of shigellosis today, in order:

1. Prevention is environmental. Clean water, sanitation, hand hygiene and food safety. This is the entire preventive arsenal and it is exactly the arsenal that is unavailable in the settings where the disease kills.
2. A case is often not diagnosed to the organism. Most childhood diarrhoea in endemic settings is treated on symptoms without a stool culture, so the specific bacterium is frequently never identified.
3. Illness is treated with rehydration, oral or intravenous, which addresses the fluid loss and not the infection.
4. Antibiotics are given for dysentery or severe disease — typically ciprofloxacin or azithromycin. This is the step that is failing. Multidrug-resistant Shigella is now common enough that the World Health Organization lists a Shigella vaccine as a development priority, and antibiotic resistance is the reason the unmet need is growing rather than shrinking [VERIFIED — Valneva/LimmaTech release, 2025-04-09; WHO Product Development for Vaccines Advisory Committee *Shigella* session materials, read 2026-08-11].
5. Travellers get advice, not protection. DUKORAL — which Valneva itself sells — protects against cholera and against travellers’ diarrhoea caused by one type of E. coli. It does nothing against Shigella.

Where S4V2 fits. At step 1, as an entirely new preventive step that does not exist today. It displaces nothing and competes with no product. That is what makes the unmet need so high — and it also means there is no established market, no reimbursement precedent and no comparator price.

B.1 Competitive landscape and positioning

Where this asset wins, and the single fact the thesis rests on.

The competitive field is small and every entrant is pre-approval. The named programs are:

• altSonflex1-2-3 — GSK Vaccines Institute for Global Health, licensed to Bharat Biotech, which now leads Phase 3, regulatory work and manufacturing. GMMA platform (bubbles shed from the bacterial outer membrane rather than purified sugars), four serotypes. The one to beat, on timeline and on manufacturing cost at LMIC volumes [WEB ESTIMATE — GSK release on the Bharat Biotech licence, read 2026-08-11].
• ShigETEC — EveliQure Biotechnologies. An oral, live-attenuated S. flexneri 2a engineered to also express a toxin fusion from enterotoxigenic E. coli, so one product could cover two causes of travellers’ diarrhoea. Phase 2 challenge study registered at Johns Hopkins [VERIFIED — ClinicalTrials.gov NCT07049159, read 2026-08-11].
• Invaplex AR-Detox — Walter Reed Army Institute of Research, trials in the Netherlands and Zambia [VERIFIED — ClinicalTrials.gov NCT05961059].
• IVT Shigella-04 — Inventprise, first-in-human dose escalation [VERIFIED — ClinicalTrials.gov NCT07205926].
• A bivalent S. flexneri 2a–S. sonnei conjugate — Zhifei Biological, in an efficacy, immunogenicity and safety trial in Chinese children aged 6 months to 5 years across nine provincial CDC sites [VERIFIED — ClinicalTrials.gov NCT05156528]. Two serotypes rather than four, and aimed at a domestic market, but it is a real conjugate efficacy study running now.

S4V2 wins on breadth plus infant data: four serotypes, and safety and immunogenicity already generated in 472 nine-month-old infants in the exact endemic setting a Phase 3 would run in. It loses on stage and sponsorship weight: its nearest four-serotype competitor is in Phase 3 with one of the world’s largest vaccine manufacturers behind it, while S4V2’s Phase 3 is unregistered and unfunded.

The single fact the thesis rests on: that the pending challenge study shows a statistically significant reduction in shigellosis — not a numerical one. The precedent on this exact platform produced a numerical reduction and missed, and a numerical reduction is worth very little to a company that then has to fund a Phase 3 off the result.

The framework’s calibration example here — oncology first-movers in novel mechanisms showing roughly 3.2× higher peak-sales potential at about 1.8× higher development risk — is an oncology reference point and is not applied to a prophylactic vaccine.

B.2 Addressable market

Launch markets. Two, with completely different economics. (1) Endemic low- and middle-income countries, delivered through infant immunisation programmes, most plausibly with Gavi support: very high volume, very low price per dose. (2) Travellers and military personnel from high-income countries, sold through travel clinics and defence procurement: low volume, commercial price. This is the channel Valneva is already built for — it sells IXIARO and DUKORAL into exactly it, and it supplies the US Department of Defense.

B.3 Value and feasibility

B.3a Expected peak sales.

The honest starting point is that a defensible bottom-up build is not available, and the reason is specific rather than a shrug. A bottom-up build needs eligible patients × annual net price × peak penetration. The eligible population is enormous and knowable. The price is not: no Shigella vaccine has ever been sold anywhere, so there is no realised price for this product, no comparator price for this disease, and the two channels would carry prices differing by more than an order of magnitude — a Gavi-tiered infant dose against a commercial travel-clinic dose. Inventing a blended price would put a fabricated number at the centre of every scenario below.

What exists instead is one anchor from the sponsor itself: Valneva states that “the global market opportunity for a vaccine against Shigella is estimated to exceed $500 million annually” [VERIFIED — Valneva Q1 2026 results release, 2026-05-13, and the 2024-08-01 partnership release, which uses the same figure]. That is a company statement about the whole market, not about S4V2’s share of it, and it is treated as the mid-point of the band below rather than as S4V2’s revenue.

The scenarios below are therefore built top-down from that single stated market figure and a share assumption, and the method is named so the reader can discount it accordingly. Every figure is conditional on approval, assumes a launch no earlier than 2032, and excludes the low-double-digit royalty owed to LimmaTech, any milestone payments, and any public-health funding contribution.

Which input is not defensible, named explicitly: the net price per dose, in both channels. It is the reason these are top-down figures with a stated method rather than a bottom-up build, and it is the input to demand from the company at the first opportunity.

B.3b Feasibility

B.4 Product-development risk

Framing questions.

1. Does the development plan support the target product profile claims? For the S. sonnei component, yes: the challenge study directly tests protection against that serotype in humans. For the other three serotypes, no — the plan generates antibody levels for S. flexneri 2a, 3a and 6 and no protection data at all, so three quarters of the tetravalent claim rests on the assumption that an antibody level which predicts protection against S. sonnei also predicts it against the others. For the infant target label, the plan supports safety and dose selection only.
2. Will the identified risks affect the target product profile? Yes, and the mechanism is direct: the primary endpoint counts every moderate case, so a vaccine that reliably prevents severe disease but not moderate disease misses the primary while still being a clinically valuable product. That exact outcome happened to Flexyn2a. It would leave the profile intact scientifically and damaged commercially and financially, because the Phase 3 has to be funded off the headline.
3. If a risk cannot be mitigated, is the asset still differentiated from competitors? Partly. Four serotypes plus infant safety data in an endemic setting remains a real differentiator against everything except altSonflex1-2-3. Against altSonflex1-2-3 specifically, an ambiguous efficacy result leaves S4V2 behind on stage, behind on sponsor scale, and without a manufacturing cost argument.

Readout

Sources

The modelled estimate. The registry’s primary completion date for NCT06615375 is 2026-04-14. Adding the stated default lag of two to four months for database lock, unblinding and topline analysis (rule 34) gives 2026-06-14 to 2026-08-14. data/benchmarks/readout-lag.json holds no observations at all, so this uses the stated default and is tagged [UNVERIFIED — modelled, default lag] rather than benchmarked. The arithmetic is stated so it can be argued with: the input is a day-precision registry field, the lag is a framework default, and the output brackets today.

Window

Basis. Every source agrees the data should be available around now, and the company has not released them, so the question this window answers is not “when will the trial finish” but “how far past its own guidance has this slipped”. Earliest is 2026-08-13, the company’s scheduled first-half 2026 results date — announced on 2026-07-22, two days after this analysis was locked, and the first confirmed disclosure vehicle in front of the readout. Likeliest is 2026-09-30: Valneva announces trial data in dedicated releases rather than inside results statements (the VALOR topline on 2026-03-23 is the pattern), the company’s own framing puts the development decision in “H2 2026”, and two prior slips argue against expecting the front edge. Latest is 2026-12-31, the end of the H2 window the development decision is guided into; a slip past that would break the framing the company has used all year. Precision is PERIOD because no source names a day or a month for the readout itself — “Mid-2026” is a half-year, and the one day-precision date in the source table is a registry completion date, not an announcement date. Confidence is LOW, not medium: the guidance has slipped twice, the current guidance has already expired unfulfilled, and there is no scheduled announcement date to anchor on.

Disagreement. CONSISTENT. All four dated sources point at mid-2026: BPIQ’s “Mid-2026”, the company’s “expected mid-2026”, the registry’s 2026-04-14 primary completion, and the modelled 2026-06/2026-08. None contradicts another. The tension in this program is not between sources — it is between every source and the calendar, and that is a slippage finding rather than a disagreement, recorded below and in confidence rather than here.

Date slippage. Two slips across six dated statements.

Six statements produce five transitions, of which two are genuine slips and three are reiterations or narrowings. The 2026-05-13 statement is the most recent guidance of any kind, and it is now almost three months old.

Attribution

Status.

Computed, not authored: lib/clustering.mjs’s attributionFor was run over this ticker’s full pipeline and this program’s own readout.window, for bpiq_drug_id 18578, and returned {"status": "CLEAN", "conflicts": []}.

Conflicts

Note. Not required for CLEAN, and one is given anyway because this particular CLEAN is weaker than it looks and a reader should not take it at face value. This ticker’s pipeline carries two has_catalyst: true rows: this program, and VLA15 (bpiq_drug_id 13398, the Lyme disease vaccine). VLA15’s catalyst_date is null and its catalyst_date_text is “TBA”, and lib/clustering.mjs derives no date range from a row with no date, so VLA15 is dropped from the comparison before any gap is measured. The computation is behaving exactly as designed — it refuses to fabricate a window — but the consequence is that CLEAN here means “no other dated catalyst is near this one”, not “nothing else can move the stock in this window”. Pfizer is expected to file Lyme regulatory submissions during 2026, and Valneva’s own half-year results land on 2026-08-13, the first day of this readout window. The stock-direction call below is written on that basis rather than on a bare CLEAN.

Market and timing for this event

• Plain takeaway. The market is not waiting for this. Valneva trades at 7.3% of its 52-week range after losing 37% in a single session on its dominant asset’s Phase 3 result, the readout has no options market to price it, and the guidance window has already run out without an announcement. Whatever this readout is worth scientifically, the equity is currently paying very little for it — which caps the downside of a miss and is the main reason a positive surprise has room to move.
• Months to this catalyst. 0.1 months to readout.window.earliest (2026-08-13) from the 2026-08-11 lock, 1.6 months to the likeliest (2026-09-30) and 4.7 months to the latest (2026-12-31). Said plainly: readout.precision is PERIOD, so this is a four-and-a-half-month window whose front edge is two days away, not a date. The window opens immediately because the guidance period has already elapsed, not because a date has been announced.
• Expected move around this event. Not readable from an options chain, because this ticker has no options chain at all — ../company.md C.6 records zero listed expiries and zero listed strikes, confirmed by two independent sources. The bracket used instead comes from this ticker’s own reactions in ../company.md C.7: single-digit percentage moves for everything that is not the Lyme program or the revenue base, against −37% and −19% for those two. A bracket of roughly 5% to 25% in absolute terms is the honest read for a non-dominant Phase 2 readout on this equity, and it is a bracket, not a point estimate.
• Nearest comparable past reaction. 2025-11-26, +9.15% close-to-close on the VLA15 Phase 2 final booster results. It is the closest analogue available because it is a Phase 2 clinical readout on a partnered asset, and it is an imperfect one in two ways: it concerned the dominant program rather than a secondary one, which argues the S4V2 reaction should be smaller; and it landed on a day that also carried a restructuring announcement, so 9.15% is a net figure. The two larger moves in C.7 — the Lyme Phase 3 at −37% and the FDA suspension at −19% — are not comparable: one is the dominant asset, the other is the revenue base.
• Materiality. Meaningful, not dominant, as recorded in ../company.md C.2. Valneva’s stated global market for a Shigella vaccine is “in excess of $500 million annually”, against a company selling €135–150M of product a year whose share price is set by a Lyme vaccine partnered with Pfizer. The stock-direction call below is consistent with that: a genuine event, and not the equity’s main determinant.
• Date slippage. Two slips (see Readout, above): H2 2025 → December 2025 → Mid-2026, with “Mid-2026” itself now elapsed. That is the single most decision-relevant timing fact in this document.

Spot. $5.30 per ADS, read 2026-08-11 — the 2026-08-10 close in data/prices/VALN.json, matched by BPIQ’s last_price and by Yahoo’s quote on the same day. Cited from ../company.md C.1 and C.4.

Scenario prices

Expected value. $5.36, +1.2% against spot $5.30. Method: the 32% modelled probability applied to the midpoint of each range — 0.32 × $6.40 + 0.68 × $4.875 = $5.36. The sign flips inside the probability band: at the 20% lower bound the expected value is $5.18 (−2.3%), and at the 47% upper bound it is $5.59 (+5.5%). This is arithmetic, not advice, and it is not a price target. It is also the arithmetic reason the stock-direction call below is no-edge rather than a direction.

Run-up

A run-up call is written here, and the honest headline is that there is no run-up left to trade. Rule 39 withholds a run-up call only when date_confidence floors at zero, which happens when readout.precision is UNKNOWN. This program’s precision is PERIOD, so the driver scores 15 and the rule does not fire. The call is therefore recorded rather than withheld — but the window opens two days after the entry date, so every field below describes a trade that has effectively no room to run, and saying so is the point of recording it.

Entry and exit

Why T-1 and not T-5. lib/runup.mjs’s exit rules are T-5, T-2 and T-1 trading days before readout.window.earliest. With earliest at 2026-08-13, T-5 resolves to 2026-08-06 and T-2 to 2026-08-11 — both at or before the entry date, which would define an exit that happens before the position is opened. T-1 resolves to 2026-08-12, one trading day after entry, and is the only one of the three that is coherent at all. It is chosen for that reason and not because a one-day hold is a strategy.

exit is recorded as null on the prediction, correctly: resolveExit returns null because the committed price cache ends 2026-08-10, before the window opens, so there is not yet enough history to resolve the rule against. It becomes resolvable once the cache extends past 2026-08-13.

Predicted move and peak

Priority score drivers

Five of the seven are lib/runup.mjs’s scoreDriver output, transcribed rather than hand-computed; unmet_need and value_uplift are the analyst’s own reading of this document, passed through the same function. Two drivers read in the opposite direction from the other five, and each says so in its own Basis cell.

Priority score. Computed by lib/runup.mjs’s priorityScore over the seven drivers above, never hand-computed. The unrounded value is 9.97; the stored score is the rounded integer.

Priority score 10 · formula_version 1.0.0

Settlement. Left null at lock time. Settled only on an explicit user request, from a confirmed primary source, against the committed price cache — never on this document’s own initiative.

Verdict

What I would do. Watch — and specifically, watch the 2026-08-13 half-year results, not the stock.

Why. The science is a genuine coin-flip weighted against, and the share price is not the way to express a view on it. The direct precedent for this readout — the same bioconjugate platform, the same challenge model, the same academic network, the same broad endpoint definition — cut shigellosis by 30.2% and missed significance at p=0.11, and S4V2 has to roughly double that on a study of similar size. Against that, the trial selects its dose before committing to the efficacy comparison, the antibody it raises was statistically associated with protection last time, and the disease has no vaccine at all after sixty years of trying, so a win would be a genuine first. But the equity does not pay for that view cleanly: Valneva is a revenue-generating vaccine company whose price is set by a Lyme vaccine partnered with Pfizer, this program is meaningful rather than dominant, the expected value is $5.36 against a $5.30 spot with the sign flipping inside the probability band, and there is no options market to express a defined-risk view instead. The timing is the last strike against acting: the guided window has already elapsed, so the run-up — the one part of this setup that would normally be tradeable — is over before the position could be opened.

What would change this. Upward: a disclosure that the challenge study met its primary endpoint with a vaccine efficacy above roughly 50%, together with a named Phase 3 funding partner or public-health funder. Efficacy alone converts an unfunded Phase 2 asset into an unfunded Phase 3 asset, which is why the funding half is not optional. Downward: a Flexyn2a repeat — a numerical reduction that misses the primary while winning the severe-disease secondaries — which would leave the company arguing for an expensive Phase 3 off a failed headline, or a further silent slip past 2026-12-31, which would suggest the data are not being announced because of what they say.

What to watch.

• 2026-08-13 — Valneva’s first-half 2026 results, confirmed and scheduled (announced 2026-07-22). The single most informative date in front of this program. Three things to read: whether the S4V2 data are disclosed or the guidance is restated again; whether “Mid-2026” becomes a third slip with a new date attached; and whether the development-decision language (“subject to positive results for both trials, Valneva will assume responsibility for all further development”) changes at all.
• Any date — a dedicated S4V2 press release. This is how Valneva announces trial data, including the VALOR Lyme topline. Its absence since 2026-01-15 is why this analysis places likeliest after the guided period rather than inside it.
• Any date — results posted to ClinicalTrials.gov for NCT06615375 or NCT06523231. Both currently read has_results: false; the infant study completed on 2026-01-22 and has still published nothing.
• 2026, no date given — Pfizer’s Lyme disease regulatory submissions. Not this program, and the most likely thing to swamp its reaction. It is also the trigger that makes the €2.96 warrants exercisable (../company.md C.3). The Attribution section explains why the clustering computation cannot see it.
• Any date — a Phase 3 partner, or a Gavi, CEPI or Gates Foundation commitment. The readout answers whether the vaccine works. This answers whether it gets made.
• Throughout — altSonflex1-2-3 news flow from Bharat Biotech. A four-serotype competitor reaching Phase 3 endpoints first compresses S4V2’s commercial case whichever way this readout goes.

Locked prediction

• Outcome-direction (will the readout succeed on its primary endpoint?): miss — probability 32%, band 20–47% [UNVERIFIED — modelled]. The reasoning: the single most informative input is that the same platform’s only prior efficacy test, Flexyn2a, produced a 30.2% reduction in shigellosis at p=0.11 against essentially this endpoint definition, in a challenge study of similar size run by the same academic network. To clear significance with roughly 35 participants per arm against a placebo attack rate near 60%, S4V2 needs vaccine efficacy of roughly 50% or better — close to double what its predecessor delivered. Three things hold the probability well above the floor: this trial confirms its dose in Step 1 before committing Step 2 to the efficacy comparison, which Flexyn2a did not do; anti-LPS IgG was statistically associated with protection in that same failed study (p=0.0016), so the mechanism has a real correlate to build on; and a second-generation construct exists precisely because the first generation underperformed, with an expensive 120-participant challenge study committed on the strength of it. Three things hold it below a coin flip: the permissive primary counts every moderate case, which is the definition Flexyn2a missed while winning the severe-disease secondaries; a tetravalent spreads antigen across four components at undisclosed per-serotype doses; and sixty years have produced no licensed Shigella vaccine from anyone. No third party has published a probability on this event; the nearest external markers are analyst price targets on the whole company, which span Goldman Sachs $4.90 to HC Wainwright $18.00 and say nothing about this endpoint.
• Stock-direction (which way do the shares move?): no-edge — confidence low — window 2026-08-11 to 2027-01-31, basis: the program’s own readout window (earliest 2026-08-13, latest 2026-12-31, PERIOD precision) plus four weeks for the reaction and any follow-on news to settle. Timing reads readout.window, never the BPIQ period-end placeholder (rule 23). Materiality is meaningful, not dominant, per ../company.md C.2, and this call is consistent with it: a $500M-a-year stated market opportunity, a decade from launch, inside a company whose price is set by a Pfizer-partnered Lyme vaccine. no-edge agrees with the expected value rather than contradicting it — $5.36 against a $5.30 spot is +1.2%, and the sign flips to −2.3% at the bottom of the probability band and +5.5% at the top. Attribution is CLEAN but should not be read as clean: the pipeline’s other has_catalyst row, VLA15, carries no date at all, so the clustering computation drops it before measuring anything, and Pfizer’s Lyme submissions plus the 2026-08-13 half-year results both sit inside this window. That is a further reason for no-edge and for low confidence.
• Scenario prices: positive $5.80–$7.00 · miss $4.60–$5.15
• Expected value: $5.36, +1.2% against spot $5.30 (read 2026-08-11)
• Run-up: entry $5.30 on 2026-08-11, exit rule T-1 trading days before readout.window.earliest — predicted move −2% to +6%, predicted peak +1% to +35% around 2026-09 — priority score 10, formula_version 1.0.0. Recorded rather than withheld: rule 39 fires only on UNKNOWN precision and this program’s is PERIOD. The call describes a one trading day hold, because readout.window.earliest is two days after the lock date and T-5 and T-2 both resolve at or before entry. There is no tradeable run-up window left.
• Settles on the first public disclosure of Phase 2b results from the S4V2 controlled human infection study, NCT06615375. Positive means that disclosure reports a statistically significant reduction, at conventional two-sided significance as pre-specified in the trial’s own statistical analysis plan, in the number of challenged participants meeting the trial’s pre-specified shigellosis definition — severe diarrhoea; or moderate diarrhoea plus fever, one at-least-moderate constitutional or enteric symptom, or two or more vomiting episodes in 24 hours; or dysentery plus any of those — among S4V2 recipients versus placebo, during the inpatient period Day 56 to Day 65 after challenge with 1,500 CFU of S. sonnei 53G. Anything else is a miss, including a numerical reduction that does not reach significance, and including significance reached only on a secondary endpoint such as moderate-to-severe shigellosis or early antibiotic use — which is precisely the Flexyn2a outcome. A first disclosure covering only the infant study NCT06523231, which has no efficacy endpoint, does not settle this prediction; settlement waits for the challenge study’s efficacy result. Source: Valneva’s or LimmaTech’s own press release, or a peer-reviewed publication or conference presentation of the S4V03 results, cross-checked against the ClinicalTrials.gov record for NCT06615375.
• Locked: yes · Settled: no

Program data-quality flags

1. Open Targets search_entities is BLOCKED, verbatim Rate limit exceeded for client: global, across three attempts under the retry policy — the standing global throttle. Its practical cost here is nil and is stated rather than assumed: Open Targets scores human gene–disease associations, and this program’s target is a bacterial surface polysaccharide, for which it holds no validation evidence.
2. ChEMBL returns no record for either “Shigella4V” or “S4V2” (count: 0 on both). A legitimate empty for a polysaccharide-protein bioconjugate, which has no small-molecule entry. Molecular identity and off-target profile are therefore unaddressed by that connector, and the mechanism description in A.1 rests on the peer-reviewed structural characterisation instead.
3. 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 on either record. The A.5b investigator table is therefore empty, with the search recorded.
4. No independent expert voice exists in this literature. All six indexed PubMed records carry LimmaTech authors, and every recurring academic name is a co-author on a sponsor-run study. A.5b records this as an empty independent-voices table with an UNKNOWN judgement rather than promoting a conflicted name into the independent column.
5. One named expert holds a dual relationship. Kawsar R. Talaat is first author of the Flexyn2a challenge paper with LimmaTech co-authors and is separately listed on ClinicalTrials.gov as the contact for EveliQure’s competing ShigETEC Phase 2 challenge study (NCT07049159), at the same Johns Hopkins Center for Immunization Research that is a listed S4V2 site. Recorded in A.5b prose because she is not an investigator on this program’s own trial and so has no table row.
6. The registry primary-completion date has passed with no disclosure. NCT06615375’s primary_completion_date is 2026-04-14 and has_results remains false; NCT06523231 completed on 2026-01-22 and also has no posted results. The company press feed carries no S4V2 item since 2026-01-15. This is the central timing fact of this analysis and it is a finding, not a gap.
7. The S4V2 dose levels are not public. NCT06615375 describes them only as “high dose” and “low dose”, with no microgram figures, so the per-serotype antigen dose cannot be compared against Flexyn2a’s 10 µg monovalent. A.1’s third scientific risk rests on that gap and says so.
8. What changed between S4V and S4V2 has never been disclosed. Both the registry and every company release describe S4V2 only as “the second generation”. Any claim that the second generation is more immunogenic than the first is therefore an inference, and it is tagged [UNVERIFIED] wherever it appears.
9. The peak-sales scenarios in B.3a are top-down, not bottom-up, because no net price per dose is defensible for either channel. The band is anchored on the company’s own “in excess of $500 million annually” market statement plus a share assumption, and the method is named in B.3a rather than presented as a build.
10. The IP position could not be established. No patent number, claim type or expiry date was found for the S4V2 composition or for the underlying bioconjugation platform. The licence chain — GSK to LimmaTech in July 2023, LimmaTech to Valneva in August 2024 — is verified; the estate protecting it is not, and the GSK-to-LimmaTech financial terms are undisclosed, so the full royalty stack above Valneva’s revenue is unknown.
11. The “December 2025” entry in the slip sequence comes from a paywalled secondary article (RTTNews, 2025-12-17), whose text ends after its opening line. It is recorded as reported rather than as a company statement, and the slip count of two does not depend on it: the two slips are the dropped H2 2025 date and the move to a Mid-2026 / H2 2026 framing.
12. attribution.status is CLEAN for a mechanical reason. VLA15 (bpiq_drug_id 13398) carries has_catalyst: true with a null catalyst_date, so lib/clustering.mjs derives no window for it and drops it before any gap is measured. CLEAN here means “no other dated catalyst is near this one”, and the Attribution section and the stock-direction call both say so rather than presenting the status at face value.