Coarse-Grained Simulations of Surfactant Partitioning into Lipid Membranes

Coarse-grained (CG) simulations offer long time-scales at relatively low cost, making them valuable for high-throughput applications. Membrane-water partition coefficients (KMW) are particular targets of simulation efforts, as they quantify molecular hydrophobicity (the tendency of a molecule to accumulate in organic tissue) and are an important toxicity indicator. Environmental risk assessment uses such data, but requires high accuracy and reliability. In this work, we benchmark CG simulations in the new Martini 3 force field for neutral, anionic, cationic and zwitterionic surfactants, and contrast them with other computational approaches. We assign new charged bead mappings, allowing expansion of our automatic CG parameterisation tool, cg_param. An ionic surfactants are incorporated, with application of the resulting KMW to create toxicity QSARs (quantitative structure activity relationships), showcasing the combined use of experimental and simulation data. Following this, the performance of cg_param models of cationic surfactants is detailed. Secondary, tertiary and certain other quaternary ammonium species perform well. cg_param has been expanded to integrate DASH, a neural-network trained hierarchy tree which allows rapid prediction of partial charges, and expands the models’ applicability to charged ring structures. Overall, the work develops the cg_param algorithm and validates its use with a mix of new and literature experimental KMW values, helping to expand the use of CG methods in academia and industry.