SMA Research Platform

Evidence graph for Spinal Muscular Atrophy

Biology-first target discovery
Christian Fischer / Bryzant Labs
15,945Targets
453Trials
85Drugs
7Datasets
8,004Sources
221,721Claims
232,882Evidence
73,172Hypotheses

Molecule Browser

5,238,909 total across 3 layers

Three-layer molecule pipeline. Designed = curated candidates scored end-to-end (QED, BBB, Lipinski, DiffDock). Seeds = compound library with ADMET predictions (ChEMBL, ZINC22, Enamine kinase/CNS/diversity/epigenetics, MMP-optimised). Catalog = purchasable bulk catalog (Enamine liquid stock, ZINC22 g-purchasable, ChEMBL extended) used as a search space for similarity-based hit expansion.

How does Three Molecule Layers work?

Designed (54,750 molecules) — AI-generated candidates scored end-to-end. Sources: PocketXMol pocket-conditional diffusion (LIMK2 DFG-out + ATP site), MolMIM/GenMol scaffold decoration. Each row has QED, BBB-permeable flag, Lipinski-pass flag, and DiffDock confidence where available.

Seeds (999,193 seeds, 999,193 ADMET-scored) — ChEMBL bioactive series, ZINC22 lead-like subsets, Enamine focused libraries (kinase / CNS / diversity / epigenetics), and MMP-optimised series. Each seed carries admet_bbb / admet_herg / admet_cyp3a4 / admet_dili / admet_ames predictions when scored.

Catalog (4,184,966 compounds, 47,071 ChEMBL-mapped) — purchasable / readily-orderable bulk space used for similarity search and hit expansion. Schema kept minimal (SMILES + inchikey + source + MW + ref_drug).

Designed
54,750
72 targets · 40,654 BBB · 52,204 Lipinski · 383 docked
Seeds
999,193
999,193 ADMET-scored · 12 sources
Catalog
4,184,966
47,071 ChEMBL · 12 sources
Data source:sma-research/molecules/— per-batch JSONs (molmim_*.json, pocketxmol_*, genmol_*, sma_moonshot_*)
Loading molecules…
Login → Command Center