MoLeR
GT4SD — Property-Conditioned Molecular Generation
Scaffold-based molecular generation using graph neural networks.
Performance Metrics
| Metric | Value | Description |
|---|---|---|
| Tanimoto Similarity | > 0.5 | Generated molecules maintain similarity to scaffold targets |
| ESOL Improvement | > 1 M/L | Improved aqueous solubility vs target in case study |
Test Set
GT4SD case study (scaffold optimization)
Notes
Not wired into Spectra: the GT4SD service registers this algorithm as a stub that returns an error naming a missing dependency. The metrics above are from the GT4SD paper case study, in which MoLeR and the Regression Transformer both produced molecules with Tanimoto > 0.5 to target scaffolds and improved ESOL by > 1 M/L.