Analog & de novo molecule design — 100% client-side, RDKit.js (WASM) + SMILES genetic search
Seed structure
Design mode
Property filters (optional)
* QED here is a simplified heuristic drug-likeness score (trapezoidal
desirability over MW/logP/HBD/HBA/TPSA/rotatable bonds/aromatic rings),
not the original Bickerton QED algorithm. Engine: genetic search on
SMILES tokens, RDKit-validated — runs fully in your browser, nothing
is sent to a server except the optional PubChem novelty check (sends
only the SMILES string of the compound being checked, no other data).
MVP: real ChEMBL experimental targets × Open Targets disease-association targets, overlap + graph.
No STRING PPI expansion or GO/KEGG enrichment stats yet (Phase 2 — needs a server proxy, CORS on
those APIs is unconfirmed). See netpharm.html to run it.