Alchemical free energy calculations have become an essential tool in structure-based drug design, and the computation of relative binding free energies is often of special importance. Many methods exist to calculate relative binding free energies (ΔΔG_binds), including free energy perturbation (FEP) or thermodynamic integration (TI). However, significant computational costs are required to run these traditional calculations, including hundreds of nanoseconds of sampling and dozens of independent calculations. λ-Dynamics (λD) reduces these costs via the use of a dynamic λ variable, allowing free energy differences to be computed from a single simulation. Additional λ variables can also be introduced to facilitate the sampling of multiple perturbations simultaneously. However, the identification of system-specific biasing potentials is often challenging and slow, reducing λD efficiency. Our work introduces our novel method to surmount these barriers called λ-Dynamics with Bias Updated Gibbs Sampling (LaDyBUGS). LaDyBUGS maintains the ability to model multiple perturbations simultaneously and compute many relative free energy differences from a single calculation. Dynamic bias updates continuously drive LaDyBUGS to sample different λ states, without identifying biases a priori. To evaluate the performance of LaDyBUGS, ΔΔG_binds were computed for 45 ligands spanning five protein-ligand complexes. Computed ΔΔG_binds were compared to experiment and TI/MBAR to assess the accuracy, precision, and efficiency of LaDyBUGS. When compared to experiment, LaDyBUGS showed high accuracy with a root mean square error (RMSE) of 0.97 kcal/mol. When compared to TI/MBAR, LaDyBUGS showed high precision with a RMSE of 0.35 kcal/mol, when TI used 15 ns / window of sampling. LaDyBUGS was found, on average, to use 35.3–113.1 times less sampling compared to TI/MBAR while maintaining high accuracy and precision. With these high efficiency gains, hundreds of relative binding free energies can be computed with LaDyBUGS with modest computational resources, accelerating molecular design at an unprecedented pace.

Xinqiang Ding

Appearances