free-energy perturbation

Free-energy perturbation, usually shortened to FEP, answers a sharp question: if I change one molecule into a close cousin, say by adding a methyl group, does it bind the target more tightly or less? Rather than measuring two molecules separately, FEP computes the difference in binding strength between them, which is what a chemist actually wants to know before deciding what to make next.

It works by a clever trick. Inside molecular dynamics simulations, the software gradually morphs one ligand into the other through a series of unphysical intermediate states, tracking the free-energy cost at each small step both when the ligand is bound to the protein and when it is free in water. Combining these two legs through a thermodynamic cycle gives the relative binding free energy, often expressed as a difference in predicted binding affinity between the pair.

When done carefully on a congeneric series, FEP can reach an accuracy of roughly one kilocalorie per mole, good enough to prioritize which analogs to synthesize. The caveats are real: it needs a reliable bound pose to start from, demands significant computing, struggles with large chemical changes or major protein rearrangements, and depends on force-field quality. It is a precision tool for ranking close relatives, not a general affinity predictor.