Recipe — FEP Point-Mutation Algorithm

P1 recipe — verbatim pseudocode from 2007-chipot-free-energy-calculations-book §2.8.6 (Chipot), p. ~60. Use this when computing the relative free-energy effect of a single residue mutation in a protein (e.g., E1659A in STRC, A1078C disulfide design) using the dual-topology paradigm. The dual-topology approach pairs with Recipe — Soft-Core Potential for Alchemical End Points to avoid end-point catastrophes.

Pseudocode (verbatim §2.8.6)

(a) Build the topologies representative of state 0 and state 1, and establish an exclusion list to prevent atoms that are not common to 0 and 1 from interacting. (b) Generate an ensemble of configurations that are representative of the reference state, λ. (c) For each configuration, evaluate the potential energy using the reference-state Hamiltonian, U(x; λ = 0). (d) Repeat the same calculations using the Hamiltonian of the target state. (e) For each configuration, evaluate the potential energy difference using (2.48). (f) Compute the ensemble average ⟨exp{−β[U(x; λ + Δλ) − U(x; λ)]}⟩_λ, from which the free energy difference ΔA = A(λ + δλ) − A(λ) can be derived. (g) Increment λ and go to stage (b).

— [Chipot & Pohorille 2007, p. 60, §2.8.6 algorithm steps a–g]

Equation (2.48), the dual-topology Hamiltonian:

U(x; λ) = λ U₁(x) + (1 − λ) U₀(x)

where U₀ contains the reference-state side-chain interactions and U₁ contains the target-state ones. Both topologies coexist on the same backbone atoms, never seeing each other through an exclusion list. (See Single-Topology vs Dual-Topology Alchemical Paradigms.)

Default parameter choices for STRC alchemical work

Lifted from book §2.8 + Fig. 2.7 caption (argon hydration calibration, p. ~46):

ParameterDefaultSource
Number of λ-windows21 (uneven width near 0 and 1)Fig. 2.7 caption [Chipot 2007]
Equilibration per window40 psFig. 2.7 caption
Production per window400 psFig. 2.7 caption
MD time step2 fsFig. 2.7 caption
Water modelTIP3PFig. 2.7 caption
Long-range electrostaticsparticle-mesh Ewald (PME)Fig. 2.7 caption
Van der Waals cutoff10 ÅFig. 2.7 caption
Decouple electrostatic vs vdW transformationsyes§2.8.4 (“strongly recommended”)
Window-narrowing near λ→0, 1yes§2.8.5 (“end-point catastrophe” mitigation)

The book validates this protocol on argon hydration: forward (creation) ΔA = +2.11 kcal/mol; reverse (annihilation) ΔA = −2.08 kcal/mol; experiment 2.002 kcal/mol at 298 K. Hysteresis 0.03 kcal/mol — the protocol is converged at this scale of perturbation.

Per-step efficiency upgrades (apply unless reason not to)

  1. Decouple electrostatics first, then vdW (§2.8.4) — annihilate point charges in window 1, then vdW in window 2. Rationale: simultaneous decoupling causes vanishing atoms to clash with nonzero residual charge. Two narrower P(ΔU) distributions converge faster than one broad joint one.
  2. Use BAR, not exponential averaging, in step (f) — replace ⟨exp(−βΔU)⟩ with the Bennett iterative solution. See Recipe — Bennett Acceptance Ratio Estimator. Reduces variance ~2–10×.
  3. Run forward + backward at every window (“simple overlap sampling,” §2.9.1) — pairs naturally with BAR; gives free-energy estimate plus immediate hysteresis check.
  4. Hamiltonian hopping (§2.9.2) — if windows show ergodicity issues, run all λ states in parallel and attempt swap moves with Metropolis acceptance every N_sample steps.

STRC application

  • h01 phase5 — pharmacochaperone_phase5_mmpbsa.py: this script currently runs MM-PBSA at WT and E1659A endpoints. To strengthen Phase 5 from “MM-PBSA scoring” to “rigorous alchemical scoring,” wrap step (a)–(g) with NAMD/GROMACS using the topology of the docked top-3 ligand series. Use this recipe + Recipe — Soft-Core Potential for Alchemical End Points + Recipe — Bennett Acceptance Ratio Estimator together.
  • h26 phase 1d — A1078C / S1080C / S1579C cysteine designs: dual-topology FEP between native and Cys side chains. Note: cysteine side chain has more atoms than serine — the dual-topology approach makes this trivial; the single-topology approach would require disappearing-atom handling.

Limitations to flag in any script docstring

  • Step (b) “ensemble of configurations” assumes the reference-state MD is converged — not free; check ergodicity (STRC AF3 Static Pocket Blindness to Loop Dynamics is the failure mode for h01 if loop dynamics aren’t sampled).
  • The exponential average in step (f) is biased and high-variance; always upgrade to BAR.
  • Bonded-term contributions (bond lengths and angles changing during mutation) typically cancel in bound vs. free states (§2.8.4) — flag in the script if the mutation alters a sterically-constrained bond near a clash.

Connections