2007 Chipot & Pohorille — Free Energy Calculations: Theory and Applications in Chemistry and Biology

Edited monograph, 14 chapters, Springer Series in Chemical Physics 86. Provides the load-bearing methods reference for h01 (E1659A pharmacochaperone) MD/FEP/MM-PBSA/LRA work and for any binding-free-energy or PMF computation in the STRC program.

Citation

Chipot, C.; Pohorille, A. (Eds.). Free Energy Calculations: Theory and Applications in Chemistry and Biology. Springer Series in Chemical Physics, vol. 86. Springer, Berlin Heidelberg, 2007. ISBN-13 978-3-540-38447-2. xviii + 519 pp.

TL;DR

A unified treatment of free energy methods in MD/MC: FEP, thermodynamic integration (TI), histogram methods (WHAM, Wang–Landau, multicanonical), nonequilibrium work (Jarzynski/Crooks), enhanced sampling (replica exchange), the Potential Distribution Theorem, quantum corrections, and approximate methods for drug design (LIE, PBFE, MM/PBSA, LRA pKa). The book’s thesis: most modern methods reduce to a small set of foundational ideas (Kirkwood, Zwanzig, Widom, Bennett, Torrie–Valleau) — they differ in how the ensemble is sampled and how estimators are constructed. Two tables in the entire book; the value is in the algorithms and pseudocodes.

Numbers that matter

The book is methodological — most “numbers” are illustrative, not parameters to lift. The two tabulated and one calibrated quantity worth preserving:

ParameterValueUnitsSource (page/fig/table)Uncertainty
Argon hydration ΔA, FEP creation, TIP3P, NAMD, 300 K, 1 atm, 21 windows+2.11kcal/molFig. 2.7 caption (Ch. 2)hysteresis vs. annihilation 0.03
Argon hydration ΔA, FEP annihilation, same protocol−2.08kcal/molFig. 2.7 caption (Ch. 2)
Argon hydration, experimental, 298 K2.002kcal/molFig. 2.7 caption (Ch. 2)
Total simulation time for argon hydration calibration9.24nsFig. 2.7 caption (Ch. 2)
Per-window equilibration / data collection40 / 400psFig. 2.7 caption (Ch. 2)
MD time step2fsFig. 2.7 caption (Ch. 2)
van der Waals cutoff10ÅFig. 2.7 caption (Ch. 2)
ABF illustrative force std-dev (polyalanine, ξ=19 Å)σ ≈ 13 vs. mean −1(force units)Fig. 4.4 caption / §4.6.5n≈14,000 samples to reach 10% error
AspRS Asp/Asn binding ΔΔA, MDFE benchmark used to fit PBFE protein dielectric~15kcal/mol§12.6.3 (Archontis et al.)reproduced with ε≈4
Thioredoxin Asp26 ionization relaxation free energy (MDFE)−56kcal/mol§12.6.4requires ε=3 in continuum to reproduce
Dielectric constant range: protein interior (continuum models)1–4§12.6.3, §12.6.41–2 if explicit MD already samples relaxation; 4 fits mutant data
Water dielectric (high-dielectric medium, continuum)~78§12.6 (standard)physical
Aqueous LIE Coulomb coefficient β (linear-response prediction)1/2§12.5empirical adjustment common
Jorgensen extended-LIE inhibitor RMS error (factor Xa, 60 ligands)<1.0kcal/mol§12.5 (citation 102)reported result

Verbatim table contents are in Flat-Histogram Reweighting Reference Table (Table 3.1) and MFEP Two-Stage Strategies Table (Table 6.1).

Method essentials

The book is a method anthology. Per-chapter takeaways for STRC:

  • Ch. 2 (FEP): dual-topology alchemical mutation with soft-core λ-scaling avoids end-point catastrophes; SOS (simple overlap sampling) and BAR are strictly preferred over forward exponential averaging. Pseudocode (a)–(g) at §2.8.6 is verbatim canonical for h01 phase5.
  • Ch. 4 (TI, ABF): ABF accumulates the running-average mean force in bins of ξ and applies the negative as a bias; once converged the system is effectively diffusive on a flat surface. Algorithm 1 (Velocity-Verlet outer loop) and Algorithm 2 (ABF subroutine) are extracted verbatim.
  • Ch. 5 (NEW): Jarzynski identity and Crooks fluctuation theorem; BAR estimator (5.50) needs forward + reverse work distributions and a Newton–Raphson iteration. Strongly outperforms exponential and cumulant estimators except in symmetric toy systems.
  • Ch. 6 (errors): the operative diagnostic is whether Γ*₁ ⊆ Γ*₀ — the target’s important phase space must be a subset of (or coincide with) the reference’s. If not, stage with an intermediate M whose Γ*_M covers the union or sits in the overlap.
  • Ch. 12 (approximate methods): LIE = α(⟨V_vdW⟩_prot − ⟨V_vdW⟩_solv) + β(⟨V_elec⟩_prot − ⟨V_elec⟩_solv) + γ. PBFE/MM-PBSA decompose ΔA_bind into direct interaction + ligand desolvation + protein desolvation. LRA: ΔA = (q/2)(⟨V⟩_reactant + ⟨V⟩_product) — gives pKa with one MD per endpoint instead of full alchemical λ scan.

Limitations

  • Two tables in 519 pages — almost no parameter values to lift; cite the book for methods and equations, not for force-field constants.
  • Force-field parameters (TIP3P partial charges, AMBER, CHARMM) are not tabulated; cite the original force-field papers for those.
  • No quantitative benchmark on protein–ligand binding free energy accuracy beyond illustrative examples; for h01 MM-PBSA error bands cite Genheden & Ryde 2015 instead (already in pharmacochaperone).
  • Dual-topology + soft-core is described qualitatively; specific α_vdW values used in NAMD/GROMACS soft-core defaults must come from those manuals.

Relevance to STRC

  • index: phase5 alchemical scoring of E1659A pharmacochaperone candidates uses exactly the FEP point-mutation framework (§2.8.6) with dual-topology + soft-core (§2.8.5). MM-PBSA gate already in phase5 script (pharmacochaperone_phase5_mmpbsa.py) decomposes per Eq. (12.65)–(12.67). LRA (§12.3.2) is the right tool to estimate the E1659A pKa shift quantitatively from existing 100 ns MD trajectories without re-running alchemical λ.
  • index: Phase 1d AF3 designs A1078C/S1080C/S1579C disulfide cysteines need ΔΔG_dimer scoring; FEP point-mutation in dual-topology paradigm is the canonical recipe. Soft-core λ schedule mandatory because cys side chains differ in atom count from native ser/ala.
  • index: ABF (§4.6) is the recommended PMF method for the h09 self-assembly reaction coordinate (e.g., RADA16 monomer → β-sheet); replica exchange (Ch. 8) handles configurational sampling.
  • STRC Electrostatic Analysis E1659A: the LRA framework at §12.3.2 — particularly Eq. (12.40) summing per-atom charge increments times the average potential — provides a defensible literature citation for the multi-method PAE-corrected ΔΔG approach already in the electrostatics note.

Connections