RARE / gtx-102-angelman-syndrome — GTX-102 (apazunersen) for Angelman syndrome

Program analysis · bpiq_drug_id 17188 · prepared 2026-08 · USD · framework v5.5.0 · 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].

Row resolution. fetch_company_drugs for RARE returned nine rows on 2026-08-07, every one of them reproduced in ../company.md C.2. This document analyses row id 17188: drug_name “GTX-102”, indications_text “Angelman Syndrome”, stage_event_label “Phase 3 Data readout”, catalyst_date 2026-12-31, catalyst_date_text “H2 2026”, has_catalyst true, is_high_mgmt_interest true — still the only row in the nine carrying that flag. It is the only row matching this molecule; the name was not used to resolve it, the integer id was.

This is a refresh. The previous analysis was stamped framework v3.7.0 and dated 2026-08-05, two days before this one. It carried no readout, no attribution and no kol block, because none of the three existed then. Writing those three, and the run-up call that hangs off the first of them, is what this refresh exists to do. The prediction it locks supersedes RARE-17188-2026-08-05; that record stays exactly as it was written.

Glossary

Executive summary

• What it is (one sentence): A synthetic genetic medicine injected into spinal fluid every three months, designed to un-silence the healthy spare copy of the UBE3A gene that every Angelman patient already has but cannot use, so that nerve cells start making the missing protein again.
• The event and when (as disclosed): Topline results from the Phase 3 ASPIRE study. The company said on 2026-08-04 that “data from this study are expected in the September or October timeframe”. BPIQ still carries “H2 2026” with a placeholder date of 2026-12-31, which is two months later than the sponsor’s own guidance. The judged window is 2026-09-01 to 2026-11-10, precision MONTH, confidence MEDIUM — see Readout, below.
• The main reason it could work: The disease has a spare part sitting in the cell, unused, and the drug’s only job is to unlock it. In the 74-patient open-label Phase 1/2, patients approached a mean 10-point gain on the Bayley-4 Cognitive raw score at 12 months against a stated meaningful threshold of 6 points. The mechanism now has a peer-reviewed clinical publication in the class — Roche’s rugonersen Phase 1 in Nature Medicine — and three separate sponsors running pivotal trials against the same transcript.
• The main risk, and it has sharpened since the last version: Two of the three Phase 3 programmes did not choose cognition as their primary endpoint. Ionis, holding its own Phase 1/2 dataset on the same mechanism, made Bayley-4 expressive communication REVEAL’s primary and demoted cognition to a secondary. Rugonersen’s BEACON hedges with “cognition and/or expressive communication”. ASPIRE is the only one of the three betting on cognition alone. That is a checkable fact about what sophisticated competitors think is most likely to move, and it sits on top of the older risk: the open-label 10-point gain was measured with no concurrent control, on a raw score that rises with age and with repeated testing, and this company converted a large significant biomarker gain into two Phase 3 clinical misses nine months ago.
• What it means for the stock: This is the event. The October options are pricing a 41–70% move, Baker Bros. exited entirely last quarter, and the shares are 23% off their 2026-07-01 level. The call is up, with low confidence, and it comes with a caveat the clustering pass now states formally rather than in prose: attribution is CONTAMINATED. Two exact FDA decision dates — DTX401 on 2026-08-23 and UX111 on 2026-09-19 — sit inside or immediately before this readout window, so a price move around the Angelman topline cannot be attributed to it alone.

0. Program-tier coverage — CLEARED

Company-tier coverage is in ../company.md C.0 and is not repeated here.

No mandatory program-tier row reads NOT CALLED, so this program clears the gate. One row is BLOCKED and its cost is named above. Two qualifications carry over: the PubMed narrowing, which is a disclosed deviation in method rather than a missing call, and the fact that ir.ultragenyx.com remains unreachable (see ../company.md C.8).

A. Scientific & clinical assessment

A.1 Mechanism of action and scientific rationale

Angelman syndrome is caused by losing the use of one gene, UBE3A, in brain cells. The unusual part is why the loss is total. Everybody inherits two copies of UBE3A, one from each parent, but in neurons the body deliberately switches off the father’s copy — a phenomenon called genomic imprinting. It does so by transcribing a long RNA molecule from the paternal chromosome, called the UBE3A antisense transcript (UBE3A-ATS), which runs across the gene and silences it. So in a normal brain, one copy works and one is locked. In Angelman syndrome the mother’s copy is deleted or broken, and the only remaining copy is the locked one [VERIFIED — according to PubMed, CNS Drugs review, DOI 10.1007/s40263-026-01313-9, retrieved 2026-08-07].

That is the whole therapeutic idea. The healthy gene is already in the cell. It just needs unlocking. GTX-102 is an antisense oligonucleotide — a short synthetic strand of modified genetic material — designed to bind the UBE3A-ATS transcript and trigger its destruction. Remove the lock, and the paternal UBE3A copy is read again and the neuron starts making the protein it has never made.

The drug is given by lumbar puncture into the cerebrospinal fluid, because an oligonucleotide of this kind cannot cross from the bloodstream into the brain. The maintenance dose in the extension studies is 14 mg every three months, and management states the great majority of patients sit at that dose [VERIFIED — BPIQ fetch_company_earnings_transcript, Q1 2026 call 2026-05-05].

How well the target is validated. Very well, and by an unusual route: genetics rather than pharmacology. For a monogenic deletion syndrome named after the gene, the association is not in serious doubt. What can be said this session is weaker than what the previous sweep could say, because Open Targets is BLOCKED: the previous sweep resolved UBE3A to ENSG00000114062 and Angelman syndrome to MONDO_0007113 and returned the gene as a target hit against the disease query [VERIFIED — Open Targets search_entities, 2026-08-05], and that reading is two days old and was not re-obtained today — four attempts each returned Rate limit exceeded for client: global. The stronger validation does not depend on that platform at all: three separate sponsors have independently arrived at the same mechanism — Ultragenyx’s GTX-102, Ionis’s ION582 and rugonersen — and all three now have a pivotal Phase 3 trial running [VERIFIED — CT.gov NCT06617429, NCT06914609, NCT07605429, read 2026-08-07].

What is new in the science since the previous version, and it cuts both ways. Roche’s rugonersen Phase 1 trial (TANGELO, n=61, children aged 1–12) has been published in Nature Medicine: the primary safety endpoint was met, and exploratory analysis showed “a dose-dependent partial normalization of the AS-associated electroencephalogram abnormality” and “signals of clinical improvement in core AS symptom domains beyond expectation from natural history data” [VERIFIED — according to PubMed, DOI 10.1038/s41591-025-03784-7, published 2025-07-11, retrieved 2026-08-07]. That is the first peer-reviewed clinical evidence for the class, and the previous version of this document specifically recorded its absence. It supports the mechanism. It does not support this trial’s endpoint, and the phrase “beyond expectation from natural history data” is exactly the comparison ASPIRE’s sham arm exists to replace.

The exact scientific step the next readout must prove. Not that the drug un-silences UBE3A — the molecular biology is established in patient-derived neurons and in animals. It must prove that un-silencing UBE3A in children who have already developed without it for four to seventeen years produces a measurable cognitive gain over and above what a matched child receiving a fake procedure achieves in the same twelve months. That is a much narrower question than “does the mechanism work”, and it is the question every prior open-label dataset has been unable to answer.

The honest scientific risk. Four things now, where the previous version had three. The second and the fourth are the ones that matter.

1. Timing. Angelman syndrome is a neurodevelopmental disorder and the brain is built early. It is genuinely unknown how much can be recovered in a child of ten. The field’s own reviews say outcomes are “dependent on their age or genotype” and argue for newborn screening precisely because early intervention is expected to work better [VERIFIED — according to PubMed, DOI 10.1007/s40263-026-01313-9].
2. The comparator gets the same tailwinds. The Phase 1/2 gain of about 10 points on the Bayley-4 Cognitive raw score was measured open-label, against natural history. A raw score counts items passed and rises with age; and repeatedly administering the same test to the same child produces a practice effect. A sham arm in ASPIRE experiences both. The entire trial rests on management’s assertion that deletion patients gain “less than 1 point a year” untreated [VERIFIED — BPIQ fetch_company_earnings_transcript, Q1 2026 call]. If that assertion is even modestly wrong, a 10-point drug effect becomes a much smaller delta. This has still never been shown in a control group.
3. The measurement changed. ASPIRE scores the Bayley-4 “without caregiver input”, administered by an outside firm sent to each site. That is a deliberate hardening of the endpoint and it makes the Phase 1/2 number an imperfect predictor of the Phase 3 number.
4. Two competitors with data on the same mechanism did not pick this endpoint. REVEAL’s primary is Bayley-4 expressive communication without caregiver input at Week 52, with cognition listed as the first secondary — and Ionis amended REVEAL in December 2025 “following additional review of data from the ongoing Phase 1/2 trial”, so the design reflects data in hand [VERIFIED — CT.gov NCT06914609, read 2026-08-07]. BEACON’s primary is “cognition and/or expressive communication” [VERIFIED — CT.gov NCT07605429]. Three pivotal trials, three different primary endpoints, and ASPIRE is the only one whose primary is cognition alone. This is not proof that cognition will miss. It is evidence about where two informed parties thought the signal was most reliable, and it did not exist as a checkable fact when this document was last written.

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

Trials

Feasibility matrix

Resourcing sufficiency. Yes, without qualification for this readout. The trial is enrolled, dosed and complete — the registry primary completion date of 2026-07-10 has passed. What remains is database lock, unblinding and analysis, and management has said the delay between last patient out and topline is closing out an international study with sham procedures and EEG collection rather than anything unresolved. Company cash of $436M at 2026-06-30 against recomputed burn of $32.3M a month runs to roughly August 2027 (see ../company.md C.3), so the readout is funded many times over. Resourcing risk sits on what comes after a positive result — a filing, a launch and the Aurora study running to 2030 — not on the readout.

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

A.3a Target label. GTX-102 for the treatment of Angelman syndrome caused by full maternal UBE3A deletion, in patients aged 4 to 17, administered by intrathecal injection. Aurora is the vehicle for extending that label to other ages and other genotypes.

A.3b Target product profile

A.3c Strategic Go/No-Go questions. The asset is completing a pivotal Phase 3 and heading for a filing, so the pre-Phase-III / registration set applies.

A.3d Regulatory designations.

• Breakthrough Therapy designation, granted 2025-06-27. It grants intensive FDA guidance on an efficient development programme, organisational commitment involving senior managers, and eligibility for rolling review and priority review. It does not lower the evidentiary bar. [VERIFIED — BPIQ fetch_company_press_releases, 2025-06-27]
• Fast Track designation, granted to GTX-102 while it was a GeneTx asset. It grants more frequent FDA meetings and eligibility for rolling review. [VERIFIED — BPIQ fetch_company_press_releases]
• Orphan Drug designation. Grants seven years of United States market exclusivity on approval for the designated indication, tax credits on qualifying trial costs and a waived application fee. The previous version of this document recorded this as unconfirmed; it is now corroborated but not primary-sourced — Ultragenyx’s own investor-relations boilerplate lists Orphan Drug designation for GTX-102, read through a web-search rendering because ir.ultragenyx.com itself timed out. [WEB ESTIMATE — Ultragenyx investor-relations release text surfaced by web search, read 2026-08-07]
• Rare Pediatric Disease designation. The designation that makes a priority review voucher possible on approval. Same source and same limit as the line above. [WEB ESTIMATE — Ultragenyx investor-relations release text surfaced by web search, read 2026-08-07]
• European Medicines Agency: Orphan Designation and PRIME. PRIME is the EMA’s enhanced-support scheme for medicines addressing an unmet need. Same source and same limit. [WEB ESTIMATE — Ultragenyx investor-relations release text surfaced by web search, read 2026-08-07]
• A priority review voucher is therefore a plausible outcome if approved — the chief financial officer named it explicitly as upside not in the model, alongside the two vouchers already expected from UX111 and DTX401 — but it is contingent, not granted. [VERIFIED — BPIQ fetch_company_earnings_transcript Q1 2026]

A.4 Target product profile valuation matrix

A.5 Evidence quality and endpoints

Endpoints

A.5b Key opinion leaders.

Panel as of. 2026-08-07 — the date the investigator and independent-voice searches below were run.

Investigators

ASPIRE names no investigators anywhere this framework can reach, and that is the finding. get_trial_details on NCT06617429 returns all 28 sites with contacts: null and no overall official, which is normal for a trial that has stopped recruiting. search_investigators on Angelman syndrome analysed 55 trials and returned 66 named people, none of them attached to NCT06617429. The table is therefore empty, and it is empty because the registry discloses nothing, not because nobody looked.

Independent voices

Finding one took more exclusions than inclusions, and the exclusions are the more useful half. Three obvious candidates were checked and rejected as not independent, each for a named, disclosed reason:

• Laurent Servais (University of Oxford), senior author of the 2026 CNS Drugs review of emerging Angelman therapies, is principal investigator on NCT05100810, the FAST UK Angelman natural-history study [VERIFIED — CT.gov search_investigators, 2026-08-07]. The Foundation for Angelman Syndrome Therapeutics founded GeneTx and is the party the GTX-102 royalty ultimately flows toward (B.1). That is a relationship with this program’s economics, so he cannot be recorded as independent of it.
• Elizabeth Berry-Kravis (Rush) and Lynne M. Bird (Rady Children’s / UCSD) are both authors on the rugonersen Phase 1 publication and Bird is a named principal investigator on the rugonersen BEACON Phase 3 [VERIFIED — DOI 10.1038/s41591-025-03784-7; CT.gov NCT07605429, 2026-08-07]. Both institutions are also ASPIRE sites. A disclosed relationship with a direct competitor is exactly as disqualifying as one with the sponsor.
• The Erasmus MC / Columbia group behind the independent mechanism-corroboration paper the previous version of this document leaned on states in its own Methods that “ASOs used in this study were kindly provided by Roche” [VERIFIED — according to PubMed, full text of DOI 10.1038/s41598-026-41197-9, retrieved 2026-08-07]. That is a competitor relationship disclosed in the paper itself. The corroboration still stands as science; the authors are not a disinterested voice on this program’s endpoint.

The limit of that check, stated plainly. An empty conflicts array here means the two searches named above came back with nothing, not that this person has no relationships. Neither journal’s disclosure statement was reachable through the connectors available, and 02-connectors.md is explicit that the absence of a disclosure is not evidence of no conflict.

Judgement

Dissent

Empty, and deliberately so. SUPPORTIVE_CONTESTED would require a named voice disagreeing with a positive reading; there is no positive reading here to disagree with, and inventing a dissenter to fill a row would be the same failure as inventing a supporter.

B. Commercial assessment

B.0 Current treatment algorithm

What a child with Angelman syndrome receives today:

1. Diagnosis, usually in the first two years, by genetic testing that identifies the maternal UBE3A deletion or mutation. The genotype is therefore known from the start, which is why no companion diagnostic is needed.
2. Antiseizure medication for the 80–90% who develop epilepsy. Levetiracetam and clobazam are the favoured first-line agents; valproate and clonazepam work but carry more complex side effects; cannabidiol oil is reported; ketogenic and low-glycaemic-index diets are adjuncts in refractory cases [VERIFIED — according to PubMed, DOI 10.1080/14656566.2026.2685074].
3. Symptom management for sleep disturbance, gastrointestinal problems, scoliosis and behaviour.
4. Physical, occupational, speech and behavioural therapy, indefinitely.
5. Lifelong multidisciplinary care with heavy caregiver burden. Most patients never live independently.

Where GTX-102 would fit. As a new step 2, above everything symptomatic: the first disease-modifying therapy, taken alongside rather than instead of the existing care. It displaces nothing and adds a quarterly hospital procedure. That is a genuinely new line of therapy in a disease that has never had one, and the algorithm above contains no drug that changes the trajectory at all — the review that sets it out says as much, describing seizure management as “guided by retrospective cohort data, caregiver surveys, and expert opinion in the absence of randomized controlled trials”.

B.1 Competitive landscape and positioning

Where this asset wins, in plain sentences. It wins on time, and by more than the previous version could demonstrate: the registry now dates REVEAL’s primary completion to 2027-08 and BEACON’s to 2029-03, so GTX-102 answers the disease’s central question roughly a year before anyone else can. It wins on population purity: taking deletion-only patients gives the cleanest possible test, where both competitors enrol deletion and mutation patients across much wider age bands and accept more variability in exchange for a broader label. And it would win on the endpoint structure if the company’s account of the FDA agreement holds, because two independent statistical paths are worth more than one.

It loses on breadth, and it may lose on endpoint choice. If all three drugs work, ION582’s label (ages 2–50, deletion and mutation) and rugonersen’s (ages 1–50, deletion and mutation) are both bigger than a deletion-only 4–17 label, and Ultragenyx has to run Aurora until 2030 to catch up. A first-mover advantage measured in months against a label gap measured in years is not obviously the better half of that trade.

The single fact the thesis rests on. That children with full maternal UBE3A deletion, aged 4 to 17, do not improve on the Bayley-4 Cognitive raw score over twelve months without treatment. Everything — the powering, the interpretation of the open-label 10-point gain, the meaning of the sham arm — depends on it. Management asserts it plainly and repeatedly. It has not been shown in a concurrent control group, because no such group has ever been run in this disease. ASPIRE is the first time anyone will see it — and the second-best-informed party in the field looked at the same class of data and put its money on a different subtest.

B.2 Addressable market

Launch markets would be the United States, the European Union and Japan, matching where ASPIRE ran.

B.3 Value and feasibility

B.3a Expected peak sales. Built bottom-up: treated patients × annual net price × penetration. Unlike the company’s gene therapies this is a chronic therapy, so peak revenue persists rather than decaying after a prevalent backlog clears — the single biggest structural difference between this asset and DTX301. All scenarios are conditional on approval, assume a first launch no earlier than 2028, and exclude any priority review voucher. The competitive assumption is tightened from the previous version: ION582’s registry primary completion is 2027-08 and rugonersen’s is 2029-03, so meaningful ION582 competition is assumed from roughly 2029 and rugonersen competition from roughly 2031.

On the price assumption, which is the weakest input: no rare-disease chronic intrathecal antisense oligonucleotide price is quoted here from a verified source. The $400–500k band is a modelled figure and is [UNVERIFIED — modelled]. It is the input most capable of moving every number in the table.

What is not passed through, and why the two available third-party figures cannot both be right. One source describes “conservative United States-only modelling” implying $1.5–2.4B peak across Ultragenyx’s late-stage programs; another reports GlobalData’s expiry model putting GTX-102 specifically at $30M of United States revenue by 2036 [WEB ESTIMATE — both re-checked 2026-08-07]. Those differ by a factor of fifty for overlapping subject matter. Neither is used as an input and both are recorded so a reader can see the dispersion. Per rule 6, the $30M figure in particular is not passed through: a $30M peak for the only disease-modifying therapy in a 15,000-patient United States indication is not a credible output whatever model produced it.

B.3b Feasibility

B.4 Product-development risk

Framing questions.

1. Does the development plan support the target product profile claims? Yes, for the narrow label. ASPIRE is properly designed to prove a cognitive benefit in deletion-type patients aged 4–17, and that is exactly the claim it is built for. It does not support the broader claim — other genotypes, other ages — which is Aurora’s job and Aurora finishes in 2030, by which time both competitors will have read out.
2. Will the identified risks affect the target product profile? One would, decisively. If the sham arm improves materially over twelve months, the measured drug effect shrinks and the trial’s powering assumption breaks. Nothing about the drug changes; the demonstrable claim does.
3. If a risk cannot be mitigated, is the asset still differentiated from competitors? No, and this is worse than it was. A failed ASPIRE would leave GTX-102 behind two live Phase 3 programmes in a mechanism the failure had not disproven, rather than one. The downside is not merely a lost year.

Readout

Sources

The modelled estimate. ClinicalTrials.gov gives NCT06617429 a primary_completion_date of 2026-07-10 — the day the last enrolled child completed the Day 338 visit. Topline follows database lock, unblinding and analysis. 01-rules.md rule 34 sets the default lag at two to four months where no benchmark exists, and data/benchmarks/readout-lag.json is empty, so: 2026-07-10 + 2 months = 2026-09-10, and 2026-07-10 + 4 months = 2026-11-10. That is the arithmetic, and it can be argued with by disputing either input.

Window

Basis. The earliest and likeliest edges come from the company’s own 2026-08-04 guidance, which names two specific months rather than a quarter or a half — that is why precision is MONTH and not PERIOD, and it is the only reason this program can carry a run-up call at all. The latest edge extends past the company’s own wording to 2026-11-10, the top of the modelled lag band, because a sponsor guiding “September or October” one month before September has room to slip a few weeks and the conservative direction to be wrong in is late, not early. Confidence is MEDIUM rather than HIGH because the guidance is still a two-month range from a sponsor that has not committed to a day, and rather than LOW because the registry primary completion has actually passed, enrolment finished early, and ten months of BPIQ note entries show the guidance narrowing rather than slipping.

Disagreement. CONSISTENT. The three dated sources point at the same event through different instruments: the registry marks the last Day 338 visit (2026-07-10), the modelled estimate adds the standard two-to-four-month analysis lag to it (2026-09/2026-11), and the company names the months inside that band (September or October). The one apparent outlier, BPIQ’s 2026-12-31, is the documented period-end placeholder rather than a fourth opinion — it is what “H2 2026” is stored as, and 01-rules.md rule 23 is explicit that it must not be read as a December date.

Date slippage. Zero slips, and one narrowing.

Five dated statements, four transitions, zero slips: the first four reiterate the same H2 2026 guidance across roughly nine months, and the fifth tightens it rather than pushing it out. Earlier note entries track execution rather than dates — 2024-07-17 FDA alignment on Phase 3 endpoints, 2024-12-19 first patient dosed, 2025-07-31 enrolment completed, 2025-10-30 first patient dosed in Aurora.

Attribution

Status. CONTAMINATED

Conflicts

Transcribed verbatim from lib/clustering.mjs’s attributionFor, run over RARE’s full pipeline and both sibling programs for bpiq_drug_id 17188 at CATALYST_CLUSTER_MIN_MONTHS = 6.

Note. Four other has_catalyst rows on this ticker land inside or beside this program’s readout window, and two of them carry exact, publicly disclosed FDA decision dates: the DTX401 PDUFA on 2026-08-23, nine days before the window opens, and the UX111 PDUFA on 2026-09-19, squarely inside it. Neither has a program document of its own, so neither would be visible to a check that read only program.json files. Both are approval decisions on gene therapies carrying saleable priority review vouchers, and on this company’s own record an approval is worth 4–9% and a Complete Response Letter would be worth considerably more in the other direction (../company.md C.7). The scenario ranges below therefore cannot be read as the market’s verdict on Angelman alone. The options chain says the same thing independently: the 2026-10-16 expiry that spans this readout also spans both PDUFA dates, so the 41–70% implied move it brackets is pricing three answers together (../company.md C.6).

Two of the four conflicts deserve a sentence of their own, because they changed character in this refresh. DTX301 (52 days) and UX701 (0 days) are both period-end placeholders on their own side — the pair GTX-102 × DTX301 was this corpus’s worked example of an INDETERMINATE finding, two independent placeholders 181 days apart that happened to touch. They are now confirmed: true because this program’s side of each pair stopped being a guess: writing a MONTH-precision readout.window here means one side of every pair it enters now rests on a disclosed date. That is the mechanism working as designed, not a change in what DTX301 or UX701 disclosed. It is also a reminder that confirmed is a property of the pair, not of the other row.

Market and timing for this event

• Plain takeaway. This is the event the whole equity is waiting for, it lands within about four to fourteen weeks, and it is formally impossible to isolate — the clustering pass now records that as CONTAMINATED rather than leaving it as prose. The market is positioned for a large move either way: the October options are pricing 41–70%, the put open interest is heavily concentrated in deep out-of-the-money strikes, and the shares have already fallen 23% from their 2026-07-01 level with no company-specific news to explain it.
• Months to this catalyst. 0.8 months to the earliest edge of the readout window (2026-09-01) from 2026-08-07, and 2.2 months to the likeliest (2026-10-15). Read from readout.window, never from catalyst_date — BPIQ’s 2026-12-31 would say 4.8 months and would be wrong by the sponsor’s own guidance. readout.precision is MONTH, so these are month-level figures and not day-level ones.
• Expected move around this event. Bracketed at 41% to 70%, from ../company.md C.6: the 2026-10-16 $25.00 straddle sums to $10.60 on the bids and $18.10 on the asks against a $25.82 spot. The mid of 55.6% is the centre of a band, not a price, because the spread is now 52% of the mid — the chain has become less readable than at the previous version, not more, even as volume rose. Read it with the attribution caveat above: that expiry spans both PDUFA dates as well as this readout. Even so, the narrow end of the bracket is enormous by this company’s own history — C.7 records no positive clinical or regulatory day worth more than about 9% ever.
• Nearest comparable past reaction. The honest answer is still there is no good analogue on the upside and an excellent one on the downside. The only GTX-102 row in ../company.md C.7 is 2024-12-19, the first patient dosed in ASPIRE, +5.94% — a dosing milestone that says nothing about a readout. The closest genuine analogue is 2025-12-29: a Phase 3 miss in the company’s then-lead asset, −44.6% close to close, which also printed the current 52-week low days later. It is comparable in kind (lead-asset Phase 3, binary, clinical endpoint after positive earlier data) and differs in one respect that cuts toward the bull case: setrusumab’s failure arrived with no offsetting approvals, whereas a failed ASPIRE would land beside two possible gene-therapy approvals.
• Materiality. Dominant among the catalysts, as recorded in ../company.md C.2. The B.3b arithmetic supports it: the unconditional modelled value of this program alone, roughly $1.25–2.5B, brackets most or all of the recomputed enterprise value floor of $2,203M. The directional up call below is consistent with that materiality (rule 27), as is the asymmetric scenario spread.
• Date slippage. Zero slips and one narrowing — see Readout, above, for the five dated statements the count is derived from.

Spot. $25.82, the 2026-08-06 close, cited from ../company.md C.4. Unlike the previous version’s $25.81, this price is after the 2026-08-04 second-quarter release, so it has absorbed the revenue beat, the halved burn rate and the narrowed readout guidance. Note the committed price cache ends one trading day earlier, at $24.93 on 2026-08-05; both figures and both dates are recorded in C.4, and no price fetch was run for this analysis.

Scenario prices

Expected value. Applying the 50% modelled probability to the midpoint of each range — $38.00 positive, $18.50 miss — gives $28.25, which is +9.4% against the $25.82 spot. This is arithmetic, not advice, and it is not a price target. Two things are worth stating plainly about what the arithmetic contains. The ranges imply a $19.50 spread between the two midpoints, so a 5-point change in the probability moves the expected value by about $0.98 — the number is not robust and should not be treated as one. And the +9.4% is smaller than the previous version’s +13.2% entirely because the probability moved from 55% to 50%; the two ranges are unchanged.

Run-up

Entry and exit

exit is left null, deliberately. readout.window.earliest is 2026-09-01 and the committed price cache ends 2026-08-05, so there is no bar at or before the anchor to count back from. resolveExit does not detect that condition and would return 2026-07-29 — a date before the entry — which is not an exit, it is an artifact of counting back from the end of a short series. Recording null is what the schema asks for until the rule is genuinely resolvable, and it is the honest answer today.

Predicted move and peak

Priority score drivers

Transcribed from lib/runup.mjs’s scoreDriver; the two judgement drivers are the analyst’s own reading of this document.

Priority score. Priority score 24 · formula_version 1.0.0

Two things about that number, because a bare 24 is not informative on its own. It is low mostly for two reasons that are both real and both stated above: the MONTH-precision date gate cuts the combined score of the other six roughly in half, and the CONTAMINATED attribution status pushes the one negative driver to its maximum. A program with an identical scientific case, a disclosed day and a clean window would score far higher on the same evidence. And the score is a ranking number for comparing run-up candidates, not a verdict on the program — the outcome and stock calls are scored separately and neither is affected by it.

Settlement. Null at lock time. It will be settled only on an explicit request, from a confirmed primary source, against the committed price cache.

Verdict

What I would do. Watch, and treat any position as a lottery ticket sized for a total loss.

Why. The expected value sits about 9% above spot and the asymmetry is real: the mechanism now has three sponsors in Phase 3 and a peer-reviewed clinical publication behind it, the trial is well designed and cleanly executed, the population is the purest available, and the shares have already fallen 23% into the print. Against that, the central scientific assumption — that deletion-type children do not improve on a repeatedly administered raw cognitive score over twelve months without treatment — has still never been observed in a control group, this company took a large significant biomarker gain into two Phase 3 trials nine months ago and missed both clinical endpoints, and the two competing Phase 3 sponsors, each holding their own data on the same mechanism, declined to make cognition their sole primary. The miss case costs roughly three-quarters of what the positive case gains, it arrives within four to fourteen weeks, and it cannot be hedged cleanly because two exact FDA decision dates sit inside the same window — which is why attribution reads CONTAMINATED rather than clean. The direction is up; the confidence in it is low, and those two statements belong together.

What would change this. Upward: publication or conference presentation of the detailed Phase 1/2 dataset before the topline, particularly any natural-history or untreated comparison showing what deletion-type children actually do on Bayley-4 Cognitive over twelve months. That single dataset is the load-bearing assumption and it is currently an assertion on an earnings call. A clean DTX401 approval on 2026-08-23 would also help by removing one of the three answers compressed into the same window. Downward: any report of transient lower-extremity weakness in an ASPIRE patient, which would be a near-veto in a paediatric trial; a Complete Response Letter on 2026-08-23 or 2026-09-19, which would take the offsetting cushion out of the miss case; or any Ionis disclosure explaining why REVEAL demoted cognition to a secondary after reviewing its own Phase 1/2 data.

What to watch.

• 2026-08-23 — the DTX401 approval decision. It sits nine days before the readout window opens, it is one of the two confirmed conflicts in the Attribution block, and it will move the shares independently of Angelman.
• 2026-09-19 — the UX111 approval decision, which already carries one Complete Response Letter on manufacturing observations. It is inside the window, not merely near it.
• 2026-09-01 to 2026-11-10 — the ASPIRE topline itself, on the judged readout window. Read the release for three things in this order: whether Bayley-4 Cognitive hit; whether MDRI hit; and whether the company presents an MDRI-only outcome as a success. Management said explicitly on 2026-05-05 that a missed Bayley-4 with a positive MDRI would still be “a demonstration of efficacy” and that a negative Bayley-4 with a positive MDRI would not be. That distinction is where the argument will be.
• Any date before the topline — the Foundation for Angelman Syndrome Therapeutics summit or any scientific meeting where the detailed Phase 1/2 data could appear. Management said on 2026-05-05 it would probably not come before the Phase 3 topline, so its appearance would itself be a signal, and data/congresses.json carries no meeting the company has named.
• 2026Q3 13F filings, from mid-November — whether Baker Bros. Advisors, which exited a 2,772,692-share position entirely in 2026Q2, re-enters, and whether Sofinnova holds the 3,145,888 shares it built in the same quarter. Recorded in ../company.md C.5.
• Throughout — Ionis news flow on REVEAL, and any BEACON enrolment update. REVEAL’s registry primary completion is 2027-08 and BEACON’s is 2029-03, so neither threatens the 2026 readout — but a strong ION582 result compresses GTX-102’s commercial case even if ASPIRE succeeds, because Ionis’s label would be broader.

Locked prediction

• Outcome-direction (will the readout succeed on its primary endpoint?): uncertain — probability 50%, band 35–65% [UNVERIFIED — modelled]. This is down from 55% (band 40–70%) at the superseded record, and the reason is specific. For: the Phase 1/2 effect size is large relative to the stated 6-point meaningful threshold (about 10 points at 12 months in 53 patients); n=129 is well sized for this disease; the deletion-only population removes the largest source of variance; randomisation is stratified on the primary endpoint’s own baseline; the mechanism now carries a peer-reviewed Phase 1 publication and three sponsors in pivotal trials; and enrolment finished early with guidance narrowed rather than slipped. Against: the open-label comparator was natural history, not a concurrent sham; a Bayley-4 raw score rises with age and with repeated administration, and the sham arm gets both; the endpoint was hardened to exclude caregiver input after Phase 1/2; this company converted a highly significant biomarker gain into two Phase 3 clinical misses nine months ago; and — the new fact — Ionis, holding its own Phase 1/2 dataset on the same mechanism, made expressive communication REVEAL’s primary and demoted cognition to a secondary, while rugonersen’s BEACON hedged with “cognition and/or expressive communication”. Three pivotal trials, three different primaries, and ASPIRE is the only one betting on cognition alone. The band spans 50 because the load-bearing assumption — that untreated deletion patients gain under one point a year — still rests on a management assertion no control group has tested. Where a third party has published a view: Morgan Stanley raised its target to $67 from $50 on a probability-of-success approach to this trial and described a favourable asymmetric setup [WEB ESTIMATE — TipRanks, read 2026-08-05], but the more recent sell-side flow is the other way — Guggenheim cut to $35 from $43 on 2026-08-04 “on trial uncertainty” and Wells Fargo cut to $45 on 2026-08-05 [WEB ESTIMATE — Investing.com and Defense World, read 2026-08-07].
• Stock-direction (which way do the shares move?): up — confidence low — window 2026-09-01 to 2026-11-15, basis: the readout window judged above (earliest 2026-09-01, latest 2026-11-10) plus five days of digestion. Attribution is CONTAMINATED and this call must not be read as attributable to ASPIRE alone: the UX111 PDUFA (2026-09-19) falls inside this window and the DTX401 PDUFA (2026-08-23) falls nine days before it, both on exact disclosed dates. The call is directional because ../company.md C.2 records this program as dominant among the catalysts and because the B.3b arithmetic puts its unconditional modelled value at most of the enterprise value. Confidence is low because the outcome call is now uncertain at 50%, because the miss case costs about 28% against a positive case worth about 47%, and because three events share one window.
• Scenario prices: positive $34–$42 · miss $16–$21
• Expected value: $28.25, +9.4% against spot $25.82 (arithmetic, not advice)
• Run-up: entry $25.82 on 2026-08-07, exit rule T-5 trading days before readout.window.earliest — predicted move −5% to +20%, predicted peak +5% to +30% around 2026-09 — priority score 24, formula_version 1.0.0
• Settles on — Ultragenyx publicly reporting topline results from the Phase 3 ASPIRE study (NCT06617429). Positive means the topline release states that ASPIRE achieved statistical significance versus sham on its primary endpoint, change from baseline in Bayley-4 Cognitive raw score without caregiver input at Day 338. Miss means the release reports that endpoint as not statistically significant, or the study is terminated without reporting it. This definition is deliberately narrow and scores an MDRI-only success as a miss, even though management has said on the record that it would still regard such a study as a demonstration of efficacy. The narrowness is the point: it is the only version of the question a third party can score the same way, and it is the endpoint ClinicalTrials.gov lists as the sole primary. The stock-direction call is scored separately and would very likely diverge in that scenario. Primary source: the Ultragenyx press release and the ClinicalTrials.gov record for NCT06617429.
• Supersedes RARE-17188-2026-08-05
• Locked: yes · Settled: no

Program data-quality flags

• Open Targets is BLOCKED on this analysis. Four attempts, each returning the verbatim Rate limit exceeded for client: global. The catalogued UBE3A–Angelman association was therefore not re-verified today; the 2026-08-05 reading is cited as history. Nothing in this document depends on it — three sponsors running Phase 3 trials against the same transcript is stronger evidence of target validity than a database entry — but the gap is named rather than smoothed over.
• The primary-endpoint structure is a source conflict and is reported rather than resolved. ClinicalTrials.gov NCT06617429 lists one primary outcome — change from baseline in Bayley-4 Cognitive raw score without caregiver input at Day 338 — and lists MDRI net response as the first secondary outcome. Ultragenyx’s chief executive said on the 2026-05-05 earnings call that “we have essentially two ways to succeed”, that “we negotiated this position with the FDA”, and that a missed Bayley-4 with a positive MDRI “is still a demonstration of efficacy”. The registry does not corroborate that. Both accounts are reported. The locked prediction settles on the registry-listed primary endpoint because that is the version a third party can score.
• Three Phase 3 trials, three different primary endpoints — reported, not reconciled. ASPIRE: Bayley-4 cognitive raw score at Day 338. REVEAL: Bayley-4 expressive communication raw score at Week 52, cognition first secondary. BEACON: Bayley-4 cognition and/or expressive communication at Week 56. This is not a data error; it is three sponsors disagreeing about which subtest is the most reliable registrational endpoint, and it is treated here as evidence rather than resolved.
• There is still effectively no peer-reviewed publication of GTX-102. The narrow PubMed search now returns two records — one German review of Angelman precision medicine and one 2026 paediatric-neurology review — and both mention the drug in passing. Every efficacy figure in this document comes from an earnings-call script and a press release. Management said on 2026-05-05 that detailed Phase 1/2 data would probably not be presented before the Phase 3 topline. The competing rugonersen programme now has a full Phase 1 clinical publication in Nature Medicine.
• The PubMed sweep was narrowed and that is disclosed. The broad mechanism query returned 30 records, above the ≤15 threshold at which the connector rule requires metadata on every hit. Metadata was retrieved on six, plus both hits from the narrow query. A documented deviation in method, not full coverage of the 30.
• The “independent” mechanism corroboration is not independent of the field’s commercial interests. The Erasmus MC / Columbia xenotransplantation paper the previous version cited as independent states in its own Methods: “ASOs used in this study were kindly provided by Roche”. The science stands; the independence claim needed correcting, and this document corrects it.
• ASPIRE discloses no investigators. CT.gov returns all 28 sites with contacts: null and names no overall official, and a 55-trial investigator search across Angelman syndrome returned nobody attached to NCT06617429. The KOL investigator table is therefore empty. That is a disclosure gap in the registry, not an omission here.
• The one independent voice recorded carries no reachable conflict statement. Neither of Debopam Samanta’s two 2026 reviews is in PubMed Central, so no full text could be searched for a conflict-of-interest declaration, and get_article_metadata does not expose one. The empty conflicts array records what two named searches found, and 02-connectors.md is explicit that the absence of a disclosure is not evidence of no conflict.
• A drug-name collision exists in ClinicalTrials.gov. NCT05531890 is a Grace Therapeutics betamethasone bioavailability study in ataxia telangiectasia whose intervention arms are also labelled “GTX-102”. Entirely unrelated compound, excluded. This is the CT.gov analogue of the dirty-drug-name problem 02-connectors.md documents for BPIQ.
• Aurora’s phase disagrees between sources. ClinicalTrials.gov NCT07157254 lists it as Phase 2; Ultragenyx describes it as “Aurora Ph2/3” in the BPIQ note and on the earnings call. Reported, not reconciled. It matters because a Phase 2 basket study cannot by itself support a label expansion on the timeline the company implies.
• MVX-220’s phase disagrees between sources too. ClinicalTrials.gov NCT07181837 lists Phase 1/2; 2026 peer-reviewed reviews describe it as the first gene replacement in pivotal (Phase 2/3) testing. Reported, not reconciled. It affects only how quickly a fourth competitor could arrive.
• BPIQ’s catalyst date is two months behind the company’s own guidance. BPIQ carries catalyst_date_text “H2 2026” with catalyst_date 2026-12-31; the company guides to September or October. Treating 2026-12-31 as a disclosed date would put the event two months late, and rule 23 forbids using it for timing at all. It survives in this document only as one row of readout.sources[].
• Every peak-sales figure in B.3a is modelled and the price input is the weakest link. No verified price for a chronic intrathecal rare-disease oligonucleotide was found; the $400–500k band is modelled. The two available third-party figures — roughly $1.5–2.4B across Ultragenyx’s late-stage programs, and GlobalData’s $30M for GTX-102 by 2036 — differ by a factor of fifty, neither is used as an input, and the $30M figure is explicitly rejected under rule 6.
• The deletion-genotype fraction is unverified. The roughly 65–75% figure used to size the initially addressable population in B.2 was not confirmed against a primary epidemiology source and is tagged [UNVERIFIED]. It is a direct multiplier on the low and base peak-sales scenarios.
• The regulatory-designation list is corroborated but not primary-sourced. Orphan Drug, Rare Pediatric Disease, EMA Orphan and EMA PRIME designations for GTX-102 are recorded from Ultragenyx’s own investor-relations boilerplate read through a web-search rendering, because ir.ultragenyx.com timed out after 60 seconds. They are tagged [WEB ESTIMATE] for that reason, not because the underlying facts are doubted.
• The run-up exit could not be resolved and is null. readout.window.earliest is 2026-09-01 and the committed price cache ends 2026-08-05, so resolveExit has no bar at or before the anchor to count back from and would return a date before the entry. Null is the honest record until the cache reaches the window.