This project reproduces a minimal free energy perturbation (FEP) workflow for two PIM kinase inhibitors using OpenFE / OpenMM
- Protein: Human PIM1 kinase
- Ligands:
- Compound 8 (PIM447 / LGH447) — PDB 5DWR
- Compound 3 / 5c — PDB 4N70 (same compound, renumbered between 2013–2015 papers)
- Experimental Ki (PIM1):
- 5c/3: 1 pM
- 8: 6 pM
→ ~6× tighter binding for 5c/3
→ ΔΔG (exp, 8→3) ≈ −1.06 kcal mol⁻¹
- Software: OpenFE 1.7 + OpenMM 8.2
- Protocol: Minimal RBFE (12 λ per leg, 250 ps prod / 50 ps eq per λ)
- Legs: complex (Protein + Ligand) and solvent (water box)
- Runs: 6 replicates per leg
- Temperature: 300 K Pressure: 1 bar Solvent: explicit TIP3P
| Leg | ⟨ΔG⟩ (kcal mol⁻¹) | SEM |
|---|---|---|
| Complex | −24.25 | 0.08 |
| Solvent | −22.77 | 0.07 |
| ΔΔG(8 → 3) | −1.48 ± 0.11 | — |
Predicted fold-change: e^(ΔΔG / RT) ≈ 12× tighter for 3/5c vs 8
Experimental fold-change: ≈ 6× tighter
Agreement: same direction, slightly larger magnitude (+0.4 kcal mol⁻¹ deviation)
- The FEP reproduced the correct trend with reasonable magnitude.
- Sign flip checks confirm proper ligand ordering (negative ΔΔG = j tighter).
- Minor overestimation likely stems from limited sampling (250 ps per λ).
rbfe_python_tutorial_two_ligands.ipynb— notebook performing setup, run, and analysis- — ligand coordinates
- — receptor coordinates
README.md— this summary
- J. Med. Chem. 2015, 58, 6599–6616 — Identification of N-(4-((1R,3S,5S)-3-amino-5-methylcyclohexyl)… (LGH447) — reports 6 pM Ki for compound 8.
- ACS Med. Chem. Lett. 2013, 4, 436–440 — Structure-Guided Optimization of Pan-PIM Inhibitors — reports 1 pM Ki for compound 5c (same as 3). """
readme_path = Path("/mnt/data/README.md") readme_path.write_text(readme_text) readme_path