Four major alchemical free energy engines, run with one shared force field, deliver statistically equivalent accuracy on a standard benchmark
- A common input-generation workflow lets four free energy engines (Amber, GROMACS, OpenMM and BLaDE) compute relative binding free energies with identical protein and ligand force fields.
- The comparison used a well-known benchmark set of congeneric ligand transformations plus two additional applications, judged against experimental measurements.
- All four methods reproduced experimental trends with broadly similar accuracy and no significant differences, so engine choice can rest on throughput, portability and perturbation network design.
💡 Points Based on the abstract
Equivalent accuracy across engines mostly says the shared force field and setup set the ceiling, not the integrator. A well-trodden benchmark is also where every engine has been tuned. A fairer test is your own congeneric series with known outcomes, including charge-changing or scaffold-hopping edges, with replicate runs compared for agreement.