The Setup

Taysha Gene Therapies is a $1.2 billion single-asset bet on TSHA-102, a gene therapy for Rett syndrome. Everything else — SLC6A1, SLC13A5, GAN, CLN1 — has been returned to UT Southwestern or discontinued. This company either gets a gene therapy approved or it goes to zero. There is no plan B.

The clinical data is the cleanest I've seen in rare disease gene therapy. Twelve out of twelve Phase 1/2 patients gained or regained developmental milestones. Zero treatment-related serious adverse events. Zero dose-limiting toxicities. The first patient is approaching three years post-dose. Management says they've "never seen a loss of any gain." Average gains per patient: roughly 20. The pivotal trial is enrolling on track, with excess screening patients, and the last of 15 patients expected to be dosed in Q2 2026.

FDA has given TSHA-102 every expedited designation that exists: Breakthrough Therapy, RMAT, Fast Track, Orphan Drug, Rare Pediatric Disease. More importantly, FDA endorsed a 6-month interim BLA pathway and a manufacturing comparability approach that lets TSHA pool data from Phase 1/2, pivotal, and ASPIRE trials — up to 30 patients in the BLA package rather than just 15. The preclinical package is complete per FDA. CMC and preclinical BLA modules are being written now, in parallel with the trial. If the 6-month interim data is clean (expected Q4 2026), TSHA could file a BLA within 60-90 days.

The stock is at $4.40, down 27% from its 52-week high. RSI 44. Short interest at 20% of float with 21.6 days to cover. Options term structure in backwardation — vol expected to increase. Fourteen analysts cover it, all at Buy, mean target $11.29. The speculative crowd has piled into December 2026 $10 calls (1,583 contracts of open interest), while institutional money is buying $8 puts (100 contracts traded the day we looked). Both sides are positioning for a big move.

The Competitor Can't Follow Them

Neurogene's NGN-401 is the only other gene therapy in registrational development for Rett syndrome. On paper, they're running neck-and-neck: same dosing completion timeline (Q2 2026), same Breakthrough Therapy designation (NGNE got theirs in February 2026), comparable clinical data (8 patients, 35 milestones gained).

But FDA explicitly told Neurogene — in writing — that a 6-month endpoint "may not be considered clinically meaningful" for NGN-401. Neurogene removed the 6-month interim analysis from their Embolden trial. Their primary endpoint is at 12 months: composite responder (CGI-I of 3 or better plus at least one developmental milestone gain), with a success threshold of 7 out of 20 patients (35%).

This isn't a subjective call. The mechanism matters. TSHA-102 uses a self-complementary AAV9 construct (scAAV), which skips the rate-limiting second-strand synthesis step and produces therapeutic protein faster. Neurogene's NGN-401 uses a single-stranded AAV9 (ssAAV) carrying the full-length MECP2 gene — it literally can't fit in a self-complementary package because full-length MECP2 exceeds the ≈2.3 kb capacity limit. TSHA-102 uses a miniMECP2 construct plus miRARE autoregulatory technology that fits within scAAV's packaging constraints. The tradeoff is truncated gene versus full-length, but the onset difference is real and FDA's differential treatment of the two programs reflects it.

We confirmed the 12-month endpoint constraint from Neurogene's own SEC filings. Their 10-Q (filed November 2025) describes the Embolden trial design. The Embolden trial design press release (June 2025) quotes FDA: "a 6-month endpoint may not be considered clinically meaningful." Coverage from the Guggenheim Summit (February 2026) states FDA's position in writing. This is not TSHA management spin.

TSHA's CSO Suku went further on the Q4 call: "I have not seen any data from Neurogene's initial studies that show actual clinical efficacy in first six months post-dosing." That's overstated. Neurogene's own 10-Q reports: "Those more recently dosed with six months of follow-up also gained developmental milestones/skills" and "three participants with six months of follow-up have also showed early efficacy, consistent with previously dosed participants." So there IS 6-month efficacy signal in NGNE's data — Suku was using a stricter definition of "clinical efficacy" than what NGNE's filings describe. The directional point holds (scAAV onset is faster, and FDA agrees), but the specific claim was adversarial positioning, not neutral analysis. Credibility on Suku's competitive commentary: 0.80, not 0.95.

The structural filing advantage for TSHA is 6-12 months minimum. Both companies complete dosing Q2 2026. TSHA's 6-month interim puts BLA filing at Q4 2026/Q1 2027. Neurogene's 12-month primary puts BLA filing at H2 2027 at earliest. TSHA also has the delivery route advantage: intrathecal (lumbar puncture, outpatient, any neurologist) versus NGNE's intracerebroventricular (surgical, neurosurgeon required, center of excellence only). Fifty percent of Rett patients are outside centers of excellence. That's a commercial moat, not just a clinical convenience.

The Question Nobody Asked

Now for the part that keeps this from being a clean story.

TSHA-102 is manufactured at Catalent's Maryland gene therapy facilities. The PPQ campaign — Process Performance Qualification, which produces BLA-quality lots — starts at this facility in Q2 2026. If PPQ fails, the BLA timeline slips 12+ months. Gene therapy CMC is the number one cause of Complete Response Letters in the space. Ultragenyx got two CMC-only CRLs on UX111. Rocket Pharmaceuticals got a CMC CRL on KRESLADI.

Here's the chain of causation we traced:

In 2025, Sarepta's Elevidys gene therapy for Duchenne muscular dystrophy had two patient deaths from acute liver failure. FDA added a boxed warning, issued clinical holds, and removed the non-ambulatory indication. Sarepta, which was Catalent Maryland's largest gene therapy customer, signed a supplemental agreement in August 2025 to delay Elevidys batch delivery to "2027 and beyond." They took a $193 million charge including $165 million for excess inventory.

The demand collapse hit Catalent Maryland in waves. Three hundred and fifty layoffs in August 2025 (316 at Harmans, roughly 30 at Baltimore). Seventy-seven more in November (61 Harmans, 16 Baltimore). Ninety-six more in February 2026. Five hundred and twenty-three workers gone from a gene therapy manufacturing site in seven months.

But the layoffs aren't the whole story. FDA inspected and found problems. Baltimore site: filling operators didn't sanitize gloves frequently enough, performed "rapid movements" instead of moving "slowly and deliberately," building infrastructure not maintained in good state of repair. Harmans site: six observations including missed testing procedures, environmental monitoring failures, and operator training issues. No warning letter at Maryland — yet. But Catalent's Indiana site did receive an actual FDA warning letter, which is currently blocking Scholar Rock's BLA resubmission. FDA attention to Catalent is elevated across the organization.

On the Q4 2025 earnings call, TSHA management said exactly nothing about any of this. Not a word about Catalent. Not a word about layoffs. Not a word about Form 483 citations at the exact facility where their PPQ campaign starts in two months. Sean Nolan described their CMC position as: "We could not be in a better position right now."

No analyst asked about it either. Nine analysts on the call. Zero questions about manufacturing risk at the company's sole CDMO.

Meanwhile, Neurogene manufactures in-house at their 42,000 square foot Houston facility. No CDMO dependency. On the manufacturing dimension, Neurogene has the structural advantage.

The Insider Problem

Every member of the C-suite is selling stock. CEO Nolan sold 178,000 shares at $4.72 ($840K) on January 26. CMO Nagendran sold 316,000+ shares across multiple transactions ($2.5M+) from December 2025 through January 2026. CFO Alam sold 104,000+ shares ($492K) in January-February 2026. All three received large equity awards on January 12 (CEO: 1,008,000 shares, CMO: 427,000, CFO: 359,000), so a portion of the selling is tax-driven same-day dispositions.

But nobody is buying. Not a single open-market purchase at $4.40 from the people who know this program best. They describe TSHA-102 as "transformative" and "potentially the most important medicine for patients with Rett syndrome." Their brokerage accounts say something quieter.

This is common in pre-revenue biotech where compensation is overwhelmingly equity-based. It may be noise. But when you have binary catalysts 2-3 months out and your management team is net selling, you note it as a mild bearish signal and move on.

Two Gates

The factor regression strips TSHA to its components: 77% idiosyncratic (just above the 75% threshold), 25% XBI at a beta of 1.88, negligible SPY. Idio vol of 86%. This stock moves on its own story.

The thesis reduces to two independent binary gates:

Gate A: Clinical (P = 63%). Does the pivotal trial work and does safety hold? The primary endpoint: at least 5 out of 15 patients achieve a developmental milestone gain at 6 months, against a natural history null of 6.7% spontaneous response. Phase 1/2 showed 100% response (83% high-dose at 6 months). We put the pivotal at 72%, safety holding at 88%, chained to 63%. Edge: none — fourteen analysts have the same data.

Gate B: Manufacturing (P = 70%). Does Catalent produce BLA-quality lots? Five hundred twenty-three layoffs, Form 483 citations, elevated FDA scrutiny. Edge: moderate bearish — we tracked the facility conditions across multiple sources and cross-referenced the SRRK/Catalent Indiana warning letter pattern. Most analysts aren't asking about this.

Everything downstream — regulatory pathway, commercial opportunity, competitive positioning — is substantially de-risked. FDA has endorsed the pathway at every decision point. Daybue at $391 million FY2025 revenue ($460-490M 2026 guidance) validates that payors will reimburse Rett therapies. The ASPIRE trial was narrowed to just 3 patients ages 2-4, safety-only, 3-month minimum — a "pleasant surprise" that reduces BLA complexity. The competitive moat is structural. But these two gates are the thesis. They're independent. And their chained probability, combined with regulatory execution, gives approximately 30-38% overall approval probability.

The Market Is Roughly Right

The market implies roughly 35% probability of approval and commercialization. Back-of-envelope: ($4.40 - $0.75) / ($11.29 - $0.75) = 34.6%. Our analysis, after processing the 10-K, the earnings call, corroborating every high-signal finding across related companies, and decomposing the thesis into independent probability gates, puts it at 30-38%.

The edge is approximately zero to three percentage points. This is not a thesis that shatters consensus.

Updated scenario (4-case, 18-month horizon, from $4.40):

EV: $6.55. Return: +49% over 18 months, 30% annualized raw alpha. After subtracting XBI sector contribution (beta 1.88 at 8% forward): 12.6% idiosyncratic alpha. Edge-weighted at 35% (the fraction of idio variance where we have genuine informational advantage — Catalent awareness and NGNE confirmation): 4-10% annualized.

On a stock with 86% idio vol, the standalone Sharpe is roughly 0.05-0.11. This only works in a portfolio of 20-30 similar bets where IR scales with sqrt(N).

But here's what changes the math. The alpha isn't static — it accelerates as gates pass:

The trade isn't about the current alpha. It's about positioning to capture the jump when gates pass. A 1-2% starter at $4.40 becomes a 3-4% position at $6 after clean data, and a 5-6% position at $8 after PPQ clears. You're buying the option to scale.

The Trade

Entry: $4.00-4.50 for a starter (1-2% of portfolio). Quarter-Kelly on our estimates is 4.4%, so 1-2% is well below even that. You're positioning for optionality, not expressing conviction.

Scale triggers: Q2 2026 data shows 12-month durability across all 12 patients with zero SAEs: add to 3-4%. PPQ lots successfully manufactured: add to 5-6%. At Phase 3 sizing, hedge the XBI component (beta 1.88, 25% of variance, no edge on biotech sector sentiment).

Exit triggers: First pivotal SAE — exit immediately. PPQ failure or Catalent warning letter at Maryland — exit or halve. Pivotal primary endpoint missed — exit. Neurogene 10-month data shows superior efficacy — reassess.

Timing: The entry window is now through Q2 2026. After Q2 data drops, the stock re-rates in one direction or the other. If clean, it gaps to $5.50-7.00 and the entry is gone. If bad, it drops to $1-2 and the thesis is dead. Key dates: AAN conference (April), ACMG (May), next earnings (May 14), last pivotal patient dosed (Q2), PPQ campaign start (Q2), full data update (H2 2026), 6-month interim analysis (Q4 2026).

What resolves the thesis before entry: The 10-K filed March 19, 2026 — check the manufacturing risk language versus prior year for Catalent updates, and find the Trinity term loan structure (unknown debt terms on a pre-revenue biotech with potential milestone-contingent covenants). Those close the two highest-priority open gaps.

Conviction

I don't have high conviction that TSHA is mispriced. Market has it roughly right. I have moderate conviction that the catalyst calendar creates asymmetry worth positioning for, and high conviction that the Catalent manufacturing risk is underappreciated — nine analysts on the call, zero questions about it, while we traced 523 layoffs, Form 483 citations at both sites, and a warning letter pattern at a sibling facility.

The honest assessment: a well-priced binary bet where approaching catalysts create a window. Not a thesis that the market is wrong about probability, but a thesis that the opportunity cost of being early is lower than the opportunity cost of being late.

If the data is clean and the factory works, this goes to $10-14. If either gate fails, it goes to $1-2. The market knows this. The question is whether you want to be positioned before the answer arrives, sized to survive being wrong.

Evidence