SMA Research Platform

ROCK-LIMK2-CFL2 Pathway

INTERACTIVE

11 nodes · live claim counts

Interactive visualization of the ROCK-LIMK2-CFL2 therapeutic axis — the platform’s highest-confidence mechanistic finding. SMN protein loss triggers cytoskeletal stress: ROCK2 expression shows complex regulation in SMA (meta-analysis: log2FC −0.254 across 5 contrasts). Therapeutic relevance is via pathway activity (ROCK-LIMK-CFL cascade), not expression level. LIMK2 expression changes in SMA motor neurons — direction under investigation (original +2.81× retracted, see retractions page). CFL2 upregulated in SMA motor neurons (confirmed biomarker, direction validated across datasets). 10/14 actin genes upregulated. SMA uses LIMK2; ALS uses LIMK1. Fasudil (ROCK inhibitor) showed preclinical signal in one mouse study (Bowerman 2012), never tested in SMA clinical trials. Click any node for evidence details and live claim counts.

Absent / Depleted

Hyperactivated

Phosphorylated (inactive)

Downregulated

Dysregulated

Activated

De-repressed

Click any node to see target details, SMA status, drug candidates, and live claim counts.

Key Findings (Pipeline v2.4)

● ROCK2 is the primary target — 8× more drug_efficacy claims than LIMK2, 15× more neuroprotection evidence. Best Vina score −10.26 kcal/mol.

● ROCK2×CFL2 AF3 iPTM 0.82 — kinase-substrate encounter complex confirmed by 3-LLM red team (Claude Opus + GPT-4o + Gemini). Matches Maekawa 1999 / Bowerman 2012 literature.

● AF3 validated for transient complexes — Contreras et al. 2026 (bioRxiv) proved AF3 predicts interfaces that escape experimental capture (plant NLR, iPTM 0.76). Our approach is the same paradigm.

● De novo binder LIMK2_nb1 iPTM 0.72 — LIMK2-selective (0.72 vs 0.33 against LIMK1, Δ=0.39). First designed binder above publication threshold.

● Drug strategy: kinase inhibitors (Fasudil-class), not PPI disruptors. Target the ROCK2 catalytic cleft, not the ROCK2-CFL2 interface.

Evidence Convergence: 8 / 9 Streams

✓ GEO omics — 3 datasets (GSE290979, GSE302774, GSE87281), 3 independent labs

✓ AF3 structure prediction — ROCK2×CFL2 iPTM 0.82 (cross-validated Boltz-2 + Chai-1)

✓ Vina + AutoDock docking — 879 docks/day, ROCK2 best −10.26 kcal/mol

✓ MD simulations — 255 trajectories (10ns + 100ns), 146 stable complexes

✓ Interface analysis — 70 PPI interfaces extracted, cross-method Jaccard scoring

✓ Cross-paper synthesis — 12+ independent publications

✓ Cross-disease ALS/SMA — PFN1/PFN2 convergence, LIMK1 vs LIMK2

✓ 3-LLM Red Team — Claude Opus + GPT-4o + Gemini consensus on kinase-substrate model

○ Wet-lab validation — pending (ADP-Glo kinase assay proposed for ROCK2)

NMJ Axis Status

DOK7×MUSK (iPTM 0.40), AGRN×LRP4 (iPTM 0.38) — NMJ receptor pairs consistently weak in AF3 predictions. Full-length folds queued for Spark2 overnight (AGRN×LRP4 = 3,973 residues). HDAC6 identified as potential combination therapy target (995 compound seeds, synergy with SMN2 splicing modifiers per Brain 2026).