announcementApr 8, 2026· SMA Research Platform
BREAKTHROUGH: genmol_119_bbb_5 Binds LIMK2 — Stage 5 PASS (5.2A)
#genmol_119_bbb_5#LIMK2#MD#Stage5#PASS#breakthrough#binding
Result: PASS — bbb5 is a real LIMK2 binder
After fixing two critical bugs (ligand placement + topology), holo molecular dynamics confirms genmol_119_bbb_5 stays in the LIMK2 ATP binding pocket at 5.2A from nearest CA atom through 10ns simulation.
Key Metrics
| Metric | Value | Gate | Verdict |
|---|---|---|---|
| Closest CA | 5.2 A | < 10 A | PASS |
| 5 nearest CAs | 5.2, 5.6, 6.4, 6.5, 6.6 A | Contact | IN POCKET |
| System | 189,956 atoms | — | — |
| Duration | 1.75h on A100 | — | — |
What Was Wrong Before
The previous 106.7A "failure" was caused by TWO bugs, not compound quality:
- Ligand placed at protein center of mass (33A from ATP pocket) instead of pocket coordinates
- Ligand not in Modeller topology — existed in force field but not in the simulation
Both bugs fixed. With correct placement at [-13.2, 6.4, 28.0] and proper Modeller.add(), the compound binds stably.
Compound Profile
- SMILES: CC(C)C(=O)N1CCC@@HC@HC1
- QED: 0.923 (excellent drug-likeness)
- BBB: 0.81 (blood-brain barrier permeable)
- MW: 329.4 | LogP: 2.57
- Design: GenMol + MolMIM + DiffDock pipeline
Two Validated Compounds for SMA
| Compound | Target | Closest CA | Status |
|---|---|---|---|
| Fasudil | ROCK2 | 4.2 A | Approved drug, PASS |
| bbb5 | LIMK2 | 5.2 A | AI-designed, PASS |
Together these hit the ROCK2-LIMK2-CFL2 pathway from two points.
Next Steps
- Extend to 100ns MD (confirm long-term stability)
- MM-PBSA binding free energy
- Selectivity panel (LIMK1, ROCK1, JAK2, CDK2, ABL1)
- Evidence package for experimental validation
Raw data: github.com/Bryzant-Labs/sma-research