ROCK2-LIMK-CFL Axis · Platform Flagship
Pipeline v2.4ROCK2 kinase inhibition is the SMA platform's primary therapeutic strategy.
SMN loss triggers cytoskeletal stress via the ROCK-LIMK-CFL cascade. ROCK2 has 8× more drug_efficacy claims than LIMK2, 15× more neuroprotection evidence, and validated clinical drugs (Fasudil-class). AF3 predicts a kinase-substrate encounter complex at ROCK2×CFL2 (iPTM 0.82), confirmed by 3-LLM red team as matching Maekawa 1999 / Bowerman 2012 literature. Drug strategy: kinase inhibitors targeting the ROCK2 catalytic cleft, not PPI disruptors.
Key computational evidence
Primary targets (ROCK-LIMK-CFL axis)
| Gene | Role | UniProt | Evidence |
|---|---|---|---|
| ROCK2 | Rho-kinase (primary drug target) | O75116 | 8× drug_efficacy, Vina −10.26 |
| LIMK2 | ROCK2 substrate kinase (SMA-specific) | P53671 | Binder iPTM 0.72, selective |
| CFL2 | Cofilin-2 (biomarker, undruggable) | Q9Y281 | Upregulated in SMA MNs |
| CFL1 | Cofilin-1 (substrate) | P23528 | ROCK2×CFL2 iPTM 0.82 |
| PFN2 | Profilin-2 (MN-enriched) | P35080 | +1.22 log2FC (p=5.3e-18) |
Secondary targets (monitoring)
Pipeline v2.4 — Stage completion
| Stage | Status | Detail |
|---|---|---|
| 0 GenMol | ✅ | NIM Build, 5 API keys |
| 1 ADMET-AI | ✅ | Prestaging + Dev_india |
| 1.5 Retrosynthesis | ✅ | SA Score + QED |
| BIO-GATE | ✅ | CQO, 122 pairs approved |
| 3a Docking | ✅ | Vina 879/day + DiffDock |
| TARGET-GATE | ✅ | Blocks undruggable targets |
| DOCK-GATE | ✅ | 57% pass rate, tightened |
| 4 Selectivity | ✅ | Boltz-2 PPI cross-target |
| 4.5 Interface | ✅ | 70 analyses, Jaccard scoring |
| 5a MD 10ns | ✅ | 250+ trajectories done |
| 5b MD 100ns | ✅ | 19 real completions |
| 6 MM-PBSA | ✅ | PE timeseries scoring |
| 7 Orthogonal | ✅ | AF3 + Boltz-2 + Chai-1 |
| 7c Alanin Scan | 🔄 | 19 jobs for ROCK2×CFL2 |
| 8 Lab-in-Loop | ⬜ | Pending wet-lab validation |
Promotion criterion for wet-lab handoff
A compound qualifies for wet-lab handoff when it passes ALL pipeline stages: ADMET CNS-pass, Vina ≤ −7.0, Boltz-2 Affinity Head positive, MD 100ns RMSD < 0.6, Interface Analysis Jaccard ≥ 0.3 between AF3 and Boltz-2. Current closest: genmol_119 vs LIMK2 (4 methods, MD done). bbb_5 (genmol_119_bbb_5) in pipeline as tracer compound.
NMJ-Rescue (legacy flagship — monitoring)
The original NMJ-Rescue flagship (DOK7/MuSK/Agrin/LRP4) produced weak AF3 results (iPTM 0.38-0.40 across receptor pairs). Full-length folds are queued on Spark2 overnight scheduler. The OSF pre-registration at osf.io/fxw5j remains active — results will be published regardless of outcome by 2026-08-04. HDAC6 combination therapy identified as potential NMJ-adjacent target (995 seeds, synergy with SMN2 splicing modifiers).