“…To circumvent this sampling issue, many of the existing computational methods used to evaluate membrane permeation rely on enhanced free energy sampling techniques. − Impressively, membrane permeation has been directly observed for a few small molecules in canonical MD simulations on the nano- and microsecond time scales. For instance, Krämer et al performed unbiased MD simulations to evaluate the permeability coefficients of oxygen, water, and ethanol using counting methods and maximum likelihood estimation for the inhomogeneous solubility-diffusion (ISD) model. ,, They found that counting methods yield nearly model-free estimates for all of the three permeants, whereas the ISD model causes large uncertainties for water due to insufficient sampling and overestimates for ethanol due to collective effects in the membrane. , For larger molecules with slower permeation, however, enhanced sampling can be essential to increase the occurrence of the slowest dynamic motions that enable permeation. This can be done by biasing the potential energy surface or altering the probability density of sampled conformations. , For example, an external bias potential can be added to the Hamiltonian (as in umbrella sampling and metadynamics (MetaD)), or the system can be coupled to higher temperatures (as in replica exchange MD) to effectively reduce energy barriers and thus sample transition regions, or the transition ensemble can be selectively sampled with path sampling approaches, such as transition interface sampling or its combination with replica exchange with or without memory effects. , …”