Comparison of methods to calculate permeabilities for small molecules through a lipid bilayer

Abstract

Membrane permeability is a key parameter governing drug bioavailability. One way to quantify permeability and to investigate which physical properties may influence it is through molecular dynamics (MD) simulations. MD simulations are also an efficient approach for screening large pools of candidate molecules. To calculate permeability, the inhomogeneous solubility–diffusion (ISD) model can be used. ISD captures both thermodynamics and kinetic factors by combining values from MD-estimated free energy and diffusion coefficient profiles. These profiles can be computed using several different biasing methods differing in computational efficiency. However, there is currently no clear framework for choosing between these methods, and their reliability and convergence has not been systematically evaluated. Most strikingly, there are no publicly available tools that implement the extraction of permeability coefficients from simulation outputs. Here, we optimize permeability calculations by simulating the transition of 88 small molecules across a membrane. Three widely used sampling methods, Umbrella Sampling, Blue Moon sampling, and Accelerated Weight Histogram, are compared to determine the most efficient approach using a combination of open source and in-house tools. Permeability coefficients are obtained and analyzed for each method. The work provides practical guidelines for best practices in permeability calculations and supports their use in high-throughput drug design.