IDH1 Research Platform

Evidence graph for Spinal Muscular Atrophy

Biology-first target discovery
Christian Fischer / Bryzant Labs
46Targets
958Trials
16Drugs
0Datasets
4,188Sources
10,439Claims
10,589Evidence
1,574Hypotheses

Molecule Browser

979 molecules

Browse every AI-generated and computationally screened molecule in the SMA platform. Sources include PocketXMol pocket-conditional generation (LIMK2 DFG-out + ATP site), MolMIM/GenMol scaffold decorations, DiffDock docking results, ML-proxy virtual screening, and the ChEMBL kinase library scored through proxy docking + ADMET. Evaluate by QED (drug-likeness), BBB permeability (CNS requirement), Lipinski compliance, and DiffDock confidence. Use the filter bar to slice by target, method, or property thresholds; export the filtered set as CSV (analysis) or SDF (RDKit / PyMOL).

How does Molecule Sources work?

De novo generation via PocketXMol (pocket-conditional diffusion, LIMK2 molecules across DFG-out and ATP-site pockets), MolMIM (scaffold optimization), and GenMol (analog expansion).

Computational screening of the ChEMBL kinase library via proxy docking and ADMET prediction, tiered by multi-fidelity Bayesian posterior.

Key properties: QED (drug-likeness, ≥0.5 is good — risdiplam is ~0.55), BBB permeability (required for CNS drugs targeting spinal motor neurons), Lipinski (oral bioavailability), and DiffDock confidence (>0 = predicted binder).

979
Total
12
Targets
0
BBB Permeable
0
Lipinski Pass
0
Docked
0.819
Avg QED
Data source:sma-research/molecules/— per-batch JSONs (molmim_*.json, pocketxmol_*, genmol_*, sma_moonshot_*)
Loading molecules…