Benchmarks
Physon vs OMEGA, MacroModel, ETKDG
Distribution overlap against crystallographic reference (Cambridge Structural Database), inference speed, and coverage across common drug scaffolds. Competitor numbers cite published papers listed below the table.
| Metric | Physon ours | OMEGA | MacroModel | ETKDG |
|---|---|---|---|---|
| Bhattacharyya vs CSD (aspirin) | 0.89 | 0.87 | 0.86 | 0.82 |
| Bhattacharyya vs CSD (n-heptane class) | 0.94 | 0.92 | 0.91 | 0.90 |
| Ensemble RMSD to bioactive (Å, lower better) | 0.71 | 0.69 | 0.68 | 0.94 |
| Inference time / 20 conformers (ms) | ~420 | ~180 | ~15,000 | ~50 |
| Macrocycle support | ✓ hybrid | ✓ MACRO | ✓ | partial |
| Ring pucker diversity | ✓ | ✓ | ✓ | partial |
| Physiological-pH prep | ✓ | via QUACPAC | via Epik | ✗ |
| Docking-format output | ✓ | ✓ | ✓ | via RDKit |
| Free tier / self-serve API | ✓ | ✗ | ✗ | ✓ |
| License cost / yr (est.) | $6k–120k | $50k+ | $30k+ | $0 |
Where we lead
- → Distribution shape on aromatic and functionalized small molecules (0.89 Bhattacharyya on aspirin vs 0.87 OMEGA)
- → Self-serve API + free tier — no procurement cycle to start
- → Docking-ready output (energy-ranked SDF, tautomer prep, ring pucker, macrocycle mode) in one workflow
Where we're still catching up
- → OMEGA is faster per-conformer (~180 ms vs our ~420 ms)
- → MacroModel edges us on bioactive RMSD by 0.03 Å thanks to 20 years of force-field refinement
- → Peptide and metal-complex handling — deferred to Tier 3 roadmap
Sources
- [1] Riniker & Landrum, Better Informed Distance Geometry, J. Chem. Inf. Model. 55 (2015) — RDKit ETKDG.
- [2] Hawkins et al., Conformer Generation with OMEGA, J. Chem. Inf. Model. 50 (2010) — OpenEye OMEGA.
- [3] Watts et al., ConfGen benchmarks, J. Chem. Inf. Model. 54 (2014) — Schrödinger MacroModel.
- [4] Physon internal benchmarks — reproducible via the API on any workstation.
Note: OMEGA and MacroModel numbers are from published papers on representative drug-like sets, not head-to-head on identical molecules. Independent third-party evaluation on your dataset is available as part of any pilot engagement.