The Trade

Long TSHA at $4.59. Target 2-3% GMV, tranched: half now, half after manufacturing validation (Q2 2026). Binary outcome Q4 2026. EV $8.03 (+75%). The edge is a probability disconnect between what 12 patients showed and what the market is willing to believe about the next 15.

The Setup

Taysha bet everything on one drug. The company returned every other program to UT Southwestern and concentrated on TSHA-102 — a one-time gene therapy for Rett syndrome, a rare neurodevelopmental disorder affecting ≈6,000 diagnosed patients in the US. No approved disease-modifying therapy exists. The only approved treatment manages symptoms and does $391 million a year doing that.

Phase 1/2 dosed 12 patients. All 12 responded. High-dose cohort: 83% response at 6 months. Zero serious adverse events through 11+ months. FDA granted Breakthrough Therapy designation in September 2025. The pivotal trial is underway — 15 patients, dosing completion Q2 2026, 6-month interim readout Q4 2026. The success bar is 5 out of 15 responders.

Phase 1/2 showed 83%. The pivotal needs 33%. That is the entire thesis.

The Probability Disconnect

The stock price IS the market's probability estimate. Back it out.

$4.59 = P(success) x $11 + (1 - P) x $1.25

P(success) = 34%

Analyst consensus is $11 (14 analysts, zero sells). Failure value is ≈$1.25 (residual cash: $319.8M / 287.3M shares = $1.11). The market is pricing a one-in-three chance of success for a drug with 12/12 response against a 33% bar.

Our estimate: 72% for the pivotal endpoint alone. The full conditional chain — pivotal success (72%) x manufacturing clearance (70%) x FDA approval (85%) — gives 52%. Both well above the market's 34%.

The edge is 18 to 38 percentage points. That's not a rounding error.

Why the Market Might Be Right

Five reasons the market could be pricing 34%, and why four of them are wrong:

1. Generic binary biotech discount. Most single-asset biotechs at $4.59 with a pivotal ahead are roughly 35% likely to succeed. The base rate from Phase 1 to approval is ≈12%. The market is applying a category prior without adjusting for this specific dataset.

This stock isn't a Phase 1 hope. It has 12 patients with 100% response, 83% at 6 months, zero SAEs, Breakthrough Therapy designation, and an FDA-endorsed 6-month interim BLA pathway. The prior should be materially higher.

2. Subjective endpoint. Milestone gain across 28 defined items — functional, not a biomarker. Assessed by blinded central raters.

The natural history control is 0-6.7% spontaneous milestone gain. The Phase 1/2 signal (83%) is 12-fold above the noise floor. You'd need systematic measurement failure across blinded central raters, not random noise, to miss this.

3. Small N. 15 patients.

Run the binomial. At an 83% true response rate, P(hitting 5/15) exceeds 99.9%. Haircut the response rate by 50% — assume half the Phase 1/2 efficacy is noise, manufacturing artifact, or selection bias — and P(5/15) still exceeds 97%. You'd need the true rate to collapse below 30%, a level never observed in any TSHA-102 cohort at any time point, for even a coin flip on the success bar. The FDA set 5/15 precisely because it gives statistical power at small N. This was designed WITH the FDA, not despite them.

4. Manufacturing process change. Pivotal patients receive drug from PPQ-manufactured lots (commercial process), not Phase 1/2 lots.

This is the most legitimate concern and I'll spend a full section on it below. Gene therapy CMC is the #1 cause of complete response letters. But FDA endorsed TSHA's comparability approach, agreeing Phase 1/2 and pivotal data can be pooled. FDA doesn't pre-clear manufacturing bridges it expects to fail.

5. The market knows something we don't.

20% short interest with 21 days to cover (57.4M shares short / ≈2.7M daily volume). This is speculative shorting of a binary name, not informed positioning. Insider transactions are tax-driven RSU sales at 15-20% of awards — routine, not conviction. No insider buying, but no exodus either.

The Competitor Has a Body

TSHA's 10-K describes Neurogene (NGNE) as having "a clinical stage gene therapy program for the treatment of Rett syndrome." This understates the situation considerably.

Neurogene's NGN-401 is running a registrational trial (Embolden) on essentially the same timeline — first patient dosed Q4 2025, dosing completion Q2 2026. Both have Breakthrough Therapy designation. Both have RMAT. Regulatory designations at parity. This is a neck-and-neck race, not a clear first-mover advantage.

But one program killed a child and the other didn't.

In November 2024, the third patient receiving NGN-401's high dose died from hyperinflammatory syndrome (HLH) associated with systemic AAV exposure. The high-dose arm was discontinued. TSHA-102 has zero treatment-related SAEs across 12 patients through 11+ months.

In a rare pediatric disease with 6,000 US patients, where parents make the treatment decisions and the Rett community is small and intensely connected, that safety differential is not a nuance that gets lost in a prescribing decision.

Neurogene has one structural advantage: in-house manufacturing (TSHA depends on Catalent). But for FDA review and commercial adoption, the safety profile is likely decisive. The Rett market probably supports two gene therapies — 16,000 global patients with no disease-modifying alternative — but first-mover with the clean safety record captures the majority.

Timing note: NGNE reports earnings March 23 — four days from now. This may reveal the Embolden trial design, enrollment progress, and endpoint details. Entering a position four days before the primary competitor discloses registrational data is a real near-term risk. If NGNE reports dramatically superior efficacy at comparable safety, the thesis weakens. If NGNE data is in line or worse, the thesis strengthens. Size the first tranche accordingly.

The Manufacturing Blind Spot

This is the factor where we have no edge and must accept execution risk.

TSHA's CMO is Catalent's Maryland gene therapy facility (Harmans). Three risk signals emerged:

446 layoffs at the Maryland gene therapy site (350 in August 2025 after Sarepta pulled Elevidys demand, another 96 in early 2026). Four hundred and forty-six people let go from the specific facility making TSHA-102.

FDA sanitation citations at Maryland — failure to sanitize gloves frequently enough, workers not moving at the "slow and deliberate pace required," infrastructure deterioration.

Pattern across Catalent post-Novo acquisition. At a separate Indiana facility, FDA issued a warning letter for contamination (cat and human hair in vial stoppers). This is directly blocking Scholar Rock's BLA resubmission. Different facility, different product, but same organizational culture.

TSHA's PPQ campaign — the manufacturing validation needed for BLA — initiates Q2 2026. We estimate 70% success on first attempt.

Here's the thing the memo hasn't said yet: PPQ clearance isn't just risk to survive. It's the first underappreciated re-rating catalyst. If manufacturing validates Q2 2026, the 15% blind spot in our factor decomposition disappears, edge-weighted conviction jumps from 67% to 82%, and the market starts pricing manufacturing as resolved ahead of the pivotal readout. That's when the second tranche goes in — not because the stock is cheaper, but because the risk profile has fundamentally improved.

The Infrastructure Is Built

The Rett market isn't hypothetical. It's operating.

Acadia's Daybue — a daily oral medication that manages symptoms — hit $391M in FY2025, guides $460-490M for 2026. The revenue number is consensus. What isn't consensus is what the commercial data reveals about infrastructure readiness for a gene therapy entrant:

Patient identification is solved. 1,070 patients receiving active shipments in Q4 2025 — these families are diagnosed, engaged, and in the system.

Payor infrastructure is built. Specialty pharmacy networks, prior authorization pathways, reimbursement precedents — all established. A gene therapy launch doesn't need to build the access architecture from scratch.

Community physician adoption. 76% of new Daybue prescriptions come from community-based physicians, not specialty centers. The prescriber base has expanded beyond academic hospitals. A gene therapy administered intrathecally still requires specialty centers for the procedure, but the referral pathways exist.

Persistency signals willingness to engage. 12-month persistency above 50%, 18-month above 45% for a symptomatic-only therapy. Parents are committed to treatment.

Gene therapy pricing precedents run $2-4M per patient (Zolgensma $2.1M, Hemgenix $3.5M, Lenmeldy $4.25M). Abeona takes a high single-digit royalty on net sales — at $2M per patient and 8%, that's $160K per patient in royalty, reducing TSHA's per-patient economics to ≈$1.84M. At 1,000 patients treated, that's $1.84 billion in lifetime revenue for a company trading at $1.3 billion market cap.

Factor Decomposition

Idio Variance:  77.1%  (above 75% target)
XBI (biotech):  25.3%
SPY (market):   -2.3%
R² = 22.9%  |  σ_idio = 86.2%  |  β_XBI = 1.88

77% idiosyncratic. This stock moves on its own news. Six factors drive expected value. Edge exists in one, maybe two:

Edge-weighted: 67%. Below 75% ideal. This is a one-factor bet — clinical probability — with a manufacturing tax.

Scenarios (18 months)

Entry and Sizing

Tranche:

• Entry 1 (now, $4.59): 50% of target. Stock 24% off highs, RSI 53, between 50-day and 200-day MAs.
• Entry 2 (post-PPQ, Q2 2026): Remaining 50%. Manufacturing resolved. Edge jumps to 82%.

Kill switches:

• SAE in pivotal patients
• Catalent Maryland receives FDA warning letter
• NGNE reports materially superior data (watch March 23 earnings)
• Phase 1/2 durability deteriorates at Q2 2026 update
• Stock above $7 before PPQ (risk/reward compresses)

Options kicker: December 2026 $5 calls at $1.80. Break-even $6.80. EV 2.4x premium at our probabilities. Captures the Q4 pivotal readout with defined risk.

Size: 2.1-2.8% GMV. Bear case loss = 1.9% of portfolio.

The position is sized at 2-3% because the edge is concentrated in one factor at 67% — below the 75% threshold where we'd size full conviction. The 15% manufacturing factor is a genuine blind spot. If PPQ clears, we add. If PPQ fails, we hold and reassess.

Can I afford to be wrong? At 2.8% GMV, the 10th percentile path costs 1.9% of the portfolio. Would I take this bet 20 times? At 52-72% success probability against a 34% break-even, yes. Every time.

Would I explain this to a bookie? "I've got a drug that worked in 12 out of 12 patients, needs to work in 5 out of 15, and the stock is priced like it's a coin flip. Give me 3-to-1 on the upside."

Evidence