The SMA Research Platform has produced its first therapeutic-format nanobody designs targeting ROCK2 kinase, using mBER (Manifold Binder Engineering and Refinement) on an A100 80GB GPU.

Results: 10 trajectories, 267 VHH nanobody PDBs, best ipTM 0.6504. Top 16 binders with ipTM > 0.55 archived.

Why it matters: ROCK2 is the upstream kinase in the ROCK-LIMK2-Cofilin axis. A nanobody blocking ROCK2 could restore actin dynamics in SMA motor neurons. Nanobodies are 10x smaller than antibodies and can be engineered for BBB penetration.

Method: mBER uses AlphaFold2-Multimer backpropagation to design VHH sequences optimized for target binding. Each trajectory runs gradient descent through AF2-Multimer, evaluates via ipTM, and relaxes with OpenMM.

Limitations: ipTM 0.6504 is below the 0.7 acceptance threshold (mBER paper reports 45% experimental success at >0.7). No experimental validation yet. More trajectories needed.

Next: LIMK2 nanobody designs + increase ROCK2 trajectories to 50+.