Spirocyclic Diamines
Generated every possible shape of a drug-like molecule family. 73% had never been catalogued before.
Spirocyclic diamines are a type of ring-shaped molecule useful in drug design. I generated every possible core shape for ring sizes 4 through 8, ending up with 391 unique scaffolds and 1,381 stereoisomers. Then I checked each one against public databases: 73% were completely new (nobody had registered them). I profiled their properties to confirm they sit in the drug-like range, mapped their shapes to show they are genuinely diverse, and ran target predictions that suggest they could interact with transporters and muscarinic receptors. Published in Helvetica Chimica Acta.
Tech Stack
PythonRDKitSMARTSPandasPubChemChEMBL
Features
- Generated 391 unique scaffolds and 1,381 stereoisomers systematically.
- 73% are new, not found in PubChem.
- Confirmed all shapes sit in the drug-like property range.
- Target predictions point to transporter and muscarinic receptor activity.
- Published in a peer-reviewed journal.
Challenges
- The larger rings (7-8 membered) have very few known analogs, so predictions are less certain.
- 1,381 stereoisomers means lots of data, so I needed efficient fingerprinting to keep things manageable.
- Public databases lack close neighbors, so traditional substructure search doesn't work well.
Learnings
- Even simple molecule families can hide huge unexplored spaces.
- Shape-based tools outperform substructure search for 3D, saturated scaffolds.
- New structure doesn't mean weird properties. These stayed within useful ranges.