Accelerating potential of mean force calculations for lipid membrane permeation: System size, reaction coordinate, solute-solute distance, and cutoffs

Abstract

Molecular dynamics simulations are capable of predicting the permeability of lipid membranes for drug-like solutes, but the calculations have remained prohibitively expensive for high-throughput studies. Here, we analyze simple measures for accelerating potential of mean force (PMF) calculations of membrane permeation, namely, (i) using smaller simulation systems, (ii) simulating multiple solutes per system, and (iii) using shorter cutoffs for the Lennard-Jones interactions. We find that PMFs for membrane permeation are remarkably robust against alterations of such parameters, suggesting that accurate PMF calculations are possible at strongly reduced computational cost. In addition, we evaluated the influence of the definition of the membrane center of mass (COM), used to define the transmembrane reaction coordinate. Membrane-COM definitions based on all lipid atoms lead to artifacts due to undulations and, consequently, to PMFs dependent on membrane size. In contrast, COM definitions based on a cylinder around the solute lead to size-independent PMFs, down to systems of only 16 lipids per monolayer. In summary, compared to popular setups that simulate a single solute in a membrane of 128 lipids with a Lennard-Jones cutoff of 1.2 nm, the measures applied here yield a speedup in sampling by factor of ∼40, without reducing the accuracy of the calculated PMF.

Massively Parallel Implementation of Steered Molecular Dynamics in Tinker-HP: Comparisons of Polarizable and Non-Polarizable Simulations of Realistic Systems

“…It is widely used to scrutinize the feasibility of a reaction process. [5][6][7][8][9][10][11] To reconstruct an acceptable PMF for a given biomolecular system, Molecular Dynamics (MD) is a popular tool but it requires the use of advanced simulation packages [12][13][14][15][16][17] coupled to a well-chosen sampling strategy. Among the different techniques developed over the years, the Umbrella Sampling (US) approach 12 developed by Torrie and Valleau in 1977 and reviewed in 2011 by Kästner 18 is based on the decomposition of the reaction coordinate in several windows.…”

pVEC hydrophobic N‐terminus is critical for antibacterial activity

“…The hydrocarbon chains of phospholipids form a nonpolar core region that is unfavorable to polar and charged species . Passive diffusion through these membranes is the major entry route, but many small molecules fail to translocate into the cell because of low membrane permeability . A major breakthrough in this field is the discovery of cell‐penetrating peptides (CPPs) and protein transduction domains that can be internalized into cells with their cargos .…”

“…The uptake and membrane interaction mechanisms of some CPPs and AMPs have been examined using experimental methods for accurate representation of cellular environments and computational methods for atomistic details, but a clear description is still missing for pVEC . Peptide translocation takes place on the order of minutes, and in addition, the cellular and membrane environment is more complex than a simple membrane model.…”

“…38 The uptake and membrane interaction mechanisms of some CPPs and AMPs have been examined using experimental methods for accurate representation of cellular environments 29,[39][40][41][42] and computational methods [43][44][45][46] for atomistic details, but a clear description is still missing for pVEC. 3 Peptide translocation takes place on the order of minutes, 47,48 and in addition, the cellular and membrane environment is more complex than a simple membrane model. However, studies that use coarse-grained membrane models, 49 replica exchange simulations, 50 free energy perturbation, 51 or enhanced sampling techniques such as metadynamics, 52 umbrella sampling, [52][53][54][55] or steered molecular dynamic (SMD) simulations 56,57 have provided valuable clues about the mechanism.…”

Untitled

“…Exceptions were Arg, Lys and Asp, whose strong repulsion at z = 0 had a maximum error of 2.0−2.6 kJ even after increasing the simulation length to 250 ns. A large periodic simulation box (256 DOPC molecules, 9.4 nm box edge) was used to minimize boundary effects in the umbrella simulations (Neale & Pomes, 2016;Nitschke et al, 2016). By comparison, a 250 ns simulation performed with a small box (64 DOPC, 4.7 × 4.7 × 9 nm) containing two arginine sidechains reveals that the small simulation significantly overestimates the steepest part of the PMF at a distance of z = 1 nm from the bilayer center.…”