NMJ-Rescue · Platform Flagship
v2 LOCKED · 2026-05-06The neuromuscular junction is the SMA platform's primary research focus.
SMN restoration is now standard of care, but a measurable phenotype gap remains in Type-2/3 patients on top of nusinersen / risdiplam / onasemnogene. The platform's job, in one sentence, is to find an SMN-independent intervention that rescues a measurable SMA phenotype in patient-derived iPSC motor-neuron / myotube co-culture, on top of standard of care, within a ≤ 14-day prediction-to-result loop. NMJ-rescue is the most testable lane for that mission.
Public commitment
The platform's first OSF-pre-registered prediction is the Q3-2026 NMJ ternary fold: DOK7 × MuSK × Agrin × LRP4. It settles on internal compute on 2026-08-04 and the result is published within 30 days regardless of outcome — confirmed, refuted, or null.
See the full pre-registration on osf.io/fxw5j and the full predictions board.
Flagship targets
60 / 40 compute reallocation
| Lane | Share of total | Detail |
|---|---|---|
| NMJ flagship | 35 % | DOK7-MuSK-Agrin-LRP4-RAPSN ternary folds + STMN2 axonal partners |
| Splice / RNA-ligand ternary | 15 % | SMN2 splice machinery + risdiplam-like ternary modes |
| General PPI saturator | 10 % | Continued cross-axis scanning at lower priority |
| Compound discovery (left funnel) | 40 % | GenMol · ADMET · DiffDock · MD · Boltz-ABFE |
| Total mechanism | 60 % | Ramp to 70 % only if NMJ data justifies — no faith-allocation. |
Producer queues are weighted accordingly: the Boltz-2 NIM saturator favors A1-NMJ targets in its rotation, and the daily AF3 pipeline allocates 18 / 6 / 6 of its 30 slots to NMJ / splice / general scanning.
Kill criterion
If the 4-model consensus (AF3 · Boltz-2 · Chai-1 · NPLexer-3) on the DOK7 × MuSK × Agrin × LRP4 ternary fold does not yield a CONFIRMED status (≥ 3 models agree, all iPTM ≥ 0.65, spread < 0.30) by 2026-08-04, the flagship is publicly retired and the next-ranked Top-10 axis is promoted in its place. No silent pivot.