Docking ranks. Free energy decides
When the ranking has to hold up, alchemical free-energy perturbation computes relative binding ΔΔG across a congeneric series. It is the optional deep-validation gate — run it on the survivors when the difference between two analogs is worth the compute.
“Run an FEP screen on these six analogs against the reference ligand and rank them by ΔΔG.”
How it works
Submit a series and a reference
Provide a congeneric series and a reference ligand with known binding. FEP is designed for relative comparisons within a series, where the perturbations are small and the physics is most reliable.
Alchemical perturbation on GPU
Each analog is alchemically transformed into the reference through intermediate states, with GROMACS running the molecular dynamics. Jobs run async on GPU — this is the expensive, accurate end of the funnel.
Ranked ΔΔG returned
You receive relative binding free energies for the series, ranking analogs with a rigor docking scores cannot reach. Run it on the handful of candidates that earned it.
Proof
run_fep_screen and run_fep_check compute relative binding ΔΔG via alchemical free-energy perturbation on GROMACS.
FEP runs as asynchronous GPU jobs on the compute plane — the same Kubernetes-Jobs infrastructure that backs production molecular dynamics.
Best used on a congeneric series after docking and MD have narrowed the field — relative free energy is where ranking becomes publication-grade.
Use this when you need to
Separate two analogs that docking scores as a tie
Rank a congeneric series with free-energy rigor
Justify a final candidate selection with defensible physics
Spend deep compute only on the candidates that earned it
The rigorous answer, for the decisions that need one.
Alchemical ΔΔG on GROMACS. Optional by design — run it when the ranking has to be right.