ΔG
Rank related ligands or identify energetic residues
Use approximate end-state methods when the goal is rapid comparative analysis from equilibrated noncovalent complex trajectories.
MM/PBSA • MM/GBSA • Decomposition
MD • 06
Free-Energy Analysis
Free-Energy Analysis
& Enhanced Sampling
Learn how to select, execute, validate and report MM/PBSA, MM/GBSA, steered molecular dynamics, umbrella sampling, PMF, FEP, thermodynamic integration, BAR, MBAR, metadynamics and adaptive-bias workflows for publication-oriented molecular research.
End-state ranking
Pathway sampling
Alchemical states
Free energyReaction coordinate / collective variable
Bound
Barrier
Unbound
between states
Scientific method selection
Begin with the research question, not the software
Each free-energy approach answers a different question and makes different assumptions. The course teaches how to avoid treating all methods as interchangeable.
PMF
Measure a pathway-dependent barrier or dissociation profile
Use steered MD to define the coordinate, umbrella windows to sample it, and WHAM or MBAR to reconstruct a PMF.
SMD • Umbrella • WHAMλ
Transform one chemical state into another
Use alchemical free-energy calculations for relative or absolute state changes with carefully designed lambda schedules and overlap checks.
FEP • TI • BAR • MBARCV
Explore rare transitions and multidimensional landscapes
Use enhanced sampling when conventional MD remains trapped behind kinetic barriers and appropriate collective variables can be defined.
Metadynamics • AWH • Replica ExchangePathway A • End-state methods
MM/PBSA and MM/GBSA without black-box interpretation
Move from trajectory preparation and receptor–ligand definitions to energetic components, uncertainty, per-residue analysis and defensible scientific conclusions.
Approximate binding energetics from equilibrium snapshots
The workflow covers topology consistency, frame selection, complex/receptor/ligand grouping, implicit-solvent models, entropy options, convergence, decomposition and publication reporting. Results are interpreted as model-dependent estimates—not experimental truth.
ΔGbind ≈ Gcomplex − Greceptor − Gligand
Typical components combine molecular-mechanics energy, polar and nonpolar solvation, and—when explicitly estimated—an entropy contribution.
van der Waals
Interpret dispersion and steric packing while checking whether a few unstable frames dominate the average.
Electrostatics
Separate direct Coulombic contributions from the opposing polar-solvation response.
Polar solvation
Compare PB or GB settings, ionic strength, dielectric assumptions and sensitivity of the result.
Nonpolar solvation
Relate solvent-accessible surface terms to hydrophobic burial without overinterpreting small differences.
gmx_MMPBSAGROMACS + AMBER-based analysis
Complete gmx_MMPBSA training track
Prepare compatible trajectories and topologies, run GB/PB calculations, inspect component energies, perform per-residue decomposition and computational alanine scanning, compare entropy strategies, and generate clean tables and plots for manuscripts.
MM/GBSAMM/PBSAPer-residue decompositionAlanine scanningInteraction entropyNormal-mode entropyConvergence plotsCSV export
Pathway B • Physical pathway
From pulling trajectories to a validated PMF
A rigorous workflow for reaction-coordinate design, steered MD, umbrella-window preparation, overlap diagnosis, WHAM analysis and uncertainty estimation.
SMD → Umbrella → PMF
The pulling trajectory is a pathway generator, not automatically a converged free-energy profile. Equilibrium sampling in overlapping windows is required.
ForceTime / distance
01
Define the physical event and collective coordinate
Choose a coordinate that captures the process—distance, projection, angle, contact number, path coordinate or another physically meaningful variable.
02
Prepare a stable starting ensemble
Center, equilibrate, remove periodicity artifacts, define pull groups and confirm that the reference geometry does not create artificial torque or deformation.
03
Run and diagnose steered molecular dynamics
Inspect force-versus-time, force-versus-distance, pulling work, structural integrity and multiple pull directions or replicas when appropriate.
04
Select umbrella windows by coordinate coverage
Extract configurations across the pathway, minimize and equilibrate each window, then choose force constants that produce neighboring histogram overlap.
05
Reconstruct the PMF and estimate uncertainty
Use WHAM or MBAR, account for autocorrelation, apply bootstrap or block-based error analysis, and test the influence of discarded equilibration time.
Pathway C • Alchemical transformations
FEP, TI, BAR and MBAR across well-overlapped lambda states
Learn the thermodynamic-cycle logic behind relative and absolute free energies, along with the practical details that determine whether a result is trustworthy.
Transform interactions rather than physically removing the molecule
Alchemical methods interpolate between defined end states. The training covers topology A/B construction, charge and Lennard–Jones transformations, soft-core potentials, lambda schedules, restraints, standard-state corrections, state overlap, replicas and cycle closure.
Lambda-state designInteracting → decoupled
A12345B
Free-energy perturbation
Understand exponential averaging, directionality, overlap limitations and why bidirectional or multistate estimators are usually preferred.
Thermodynamic integration
Integrate ensemble averages of ∂H/∂λ, refine lambda spacing where curvature is high and estimate integration error.
Bennett acceptance ratio
Combine forward and reverse information between neighboring states and inspect convergence of cumulative estimates.
Multistate reweighting
Use samples across all states, calculate uncertainty and overlap, and account for correlated time-series data before analysis.
Pathway D • Rare-event sampling
Enhanced sampling for hidden states and high barriers
Choose collective variables, biasing strategies and reweighting procedures that match the biological or chemical process rather than producing decorative landscapes.
Metadynamics & well-tempered metadynamics
Deposit history-dependent bias along selected collective variables, encourage exploration of metastable states and reconstruct free-energy surfaces through appropriate reweighting.
- Collective-variable design and dimensionality control
- Gaussian height, width, pace and bias factor
- Restart-safe HILLS files and convergence monitoring
- Block analysis and free-energy uncertainty
Accelerated Weight Histogram
Use AWH for adaptive sampling of pull coordinates or alchemical dimensions while monitoring coverage, diffusion, histogram growth and disconnected sampling regions.
- Coordinate range and initial landscape setup
- Adaptive update schedule and target distribution
- Multidimensional or alchemical AWH
- Bias and free-energy output interpretation
Replica exchange & Hamiltonian exchange
Improve movement between temperature or Hamiltonian states, evaluate exchange probabilities and preserve the ensemble required for valid thermodynamic analysis.
- Temperature-REMD and HREX concepts
- Replica ladder and acceptance diagnostics
- Bias-exchange metadynamics
- Demultiplexing and state-specific analysis
Quality control
Convergence is demonstrated through diagnostics, not appearance
A smooth curve is not proof of adequate sampling. Every method receives a dedicated validation checklist and an uncertainty strategy.
Equilibration and correlation
Discard initial transient behavior and estimate effective independent samples rather than treating every saved frame as statistically independent.
Overlap and coverage
Inspect umbrella histograms, lambda overlap matrices, transition counts and unsampled coordinate regions.
Time convergence
Plot cumulative estimates, split trajectories into blocks and compare early, middle and late production intervals.
Replicates and cycle closure
Repeat sensitive calculations, compare forward/reverse paths and test thermodynamic-network consistency when multiple transformations are available.
Model sensitivity
Evaluate force field, protonation, dielectric, ionic strength, restraint and collective-variable assumptions where they materially influence conclusions.
Error bars and uncertainty language
Report statistical uncertainty separately from model-form uncertainty and avoid ranking compounds whose confidence intervals strongly overlap.
Neighboring-state overlap
Histogram and state-overlap diagnostics
Adjacent distributions should communicate enough information for stable reweighting. Poor overlap cannot be repaired by a visually attractive final curve.
Important scientific boundary: covalent complexes require different language
Conventional receptor–ligand MM/PBSA or MM/GBSA assumes a noncovalently associated complex. For a covalently bonded adduct, an end-state calculation may be framed as energetic stability or component analysis of the adduct, but it should not be reported as an ordinary noncovalent binding free energy without a properly defined chemical thermodynamic cycle.
Publication-oriented outputs
Transform simulation data into defensible scientific evidence
The training emphasizes traceable files, interpretable figures, uncertainty, methods language and results guidance rather than only command execution.
End-state energetic report
Clean component tables and comparative interpretations for noncovalent complexes.
- Total and component energy plots
- Per-residue decomposition heatmaps
- Alanine-scanning prioritization
- Frame-convergence analysis
SMD and PMF figure suite
Publication-ready pathway and energetic-barrier graphics with methodological context.
- Force vs time
- Force vs distance or position
- Umbrella histogram overlap
- PMF with confidence intervals
Alchemical free-energy package
Transparent state-wise estimates and validation summaries.
- ΔG by lambda interval
- TI integrand and BAR/MBAR estimates
- Overlap matrix and cycle closure
- Replicate and uncertainty table
Free-energy landscape analysis
Meaningful state identification from enhanced-sampling trajectories.
- 1D and 2D FES plots
- State populations and barriers
- Representative structures
- Reweighting and convergence checks
Methods and reproducibility text
Structured manuscript-ready descriptions that preserve parameter transparency.
- Software and version details
- Collective variables and restraints
- Sampling length and frame treatment
- Error-estimation protocol
Results and limitation guidance
Interpretation templates that distinguish statistical, physical and methodological claims.
- Avoid over-ranking weak differences
- Explain model assumptions
- Connect energetic terms to structure
- Propose validation experiments
Common questions
What every researcher should clarify before calculating ΔG
It is an approximate end-state framework whose result depends on sampling, force field, implicit-solvent model, dielectric assumptions and entropy treatment. It is most defensible for carefully controlled comparative analysis rather than as an unquestioned absolute experimental equivalent.
A single pulling trajectory mainly provides a nonequilibrium pathway and work/force profile. A standard umbrella-sampling PMF requires equilibrated sampling in multiple restrained windows with adequate overlap, followed by WHAM or another suitable reweighting estimator.
There is no universal fixed number. Window density should be chosen from observed overlap, curvature, endpoint behavior and the complexity of the coordinate. Adaptive refinement is usually better than copying a fixed schedule from another system.
The variables should distinguish relevant metastable states, capture slow modes and remain low-dimensional enough for practical sampling. Poor collective variables can create apparent convergence while hidden orthogonal degrees of freedom remain unsampled.
A PMF is defined along a selected coordinate or set of coordinates and therefore depends on that coordinate definition. A broader FES can be reconstructed from biased or unbiased samples across chosen collective variables, but it still reflects those variables and the sampling/reweighting strategy.
Official learning resources
Primary documentation used throughout the training workflow
Links open the official software documentation or tutorial resources in a new tab.
Free-energy parameters and estimators
Lambda states, soft-core interactions, expanded ensemble, AWH, BAR and umbrella-sampling analysis.
Open GROMACS documentation →MM/PBSA and MM/GBSA analysis
Official documentation for binding-energy calculations, decomposition, entropy and supported workflows.
Open gmx_MMPBSA docs →Umbrella-sampling PMF reconstruction
Weighted Histogram Analysis Method for GROMACS umbrella simulations, including bootstrap options.
Open gmx wham docs →Metadynamics and collective variables
Official syntax, metadynamics actions, reweighting, grids and advanced biased-sampling methods.
Open PLUMED documentation →BAR, MBAR and free-energy surfaces
Multistate reweighting, overlap, timeseries subsampling, uncertainty and FES construction.
Open PyMBAR docs →MM-PBSA thermodynamic cycle
Official tutorial covering component analysis, normal-mode entropy and computational alanine scanning.
Open AMBER tutorial →AMRA-Lab premium mentorship
Build a free-energy workflow that matches your scientific question
Receive structured guidance for system preparation, method selection, calculations, convergence checks, troubleshooting, publication-quality figures, methods writing and interpretation for protein–ligand, protein–protein, membrane, solvent and advanced sampling projects.