pymbar

pymbar implements the multistate Bennett acceptance ratio (MBAR) method to compute free energy differences and thermodynamic expectations from equilibrium samples across multiple probability densities.

Key Features:

• Multistate Estimation: Extends the Bennett acceptance ratio (BAR) to multiple thermodynamic states via the MBAR formalism to estimate free energy differences and thermodynamic expectations.
• No Histogram Binning: Computes estimates directly from equilibrium samples without discretizing energies into histograms, avoiding bias from energy binning.
• Efficiency and Reduced Complexity: Eliminates histogram reweighting and reduces computational time complexity when solving the MBAR estimating equations for large datasets.
• Statistical Uncertainty Estimation: Provides estimates of statistical uncertainties alongside free energy differences and thermodynamic expectations.
• Unbiased and Low Variance: Produces unbiased estimates with low variance in large-sample regimes relative to other methods for multistate equilibrium data.

Scientific Applications:

• Free energy landscapes in molecular dynamics: Combining equilibrium data from multiple states to analyze free energy landscapes in molecular dynamics and biophysics.
• Potential of mean force for DNA hairpins: Estimating the potential of mean force for DNA hairpins from multiple optical tweezer measurements under constant force bias.

Methodology:

Apply the multistate Bennett acceptance ratio (MBAR) algorithm to equilibrium samples across multiple probability densities to solve coupled estimating equations for free energy differences and thermodynamic expectations without histogram-based reweighting, and compute statistical uncertainty estimates.