Constructing Predictive\nImplicit Solvent Models: Insights\nfrom Exploring Potential Parameter Space and Property Predictions

Implicit\nsolvent (IS) models enhance the efficiency of\nsimulating\nliquid systems, but those based on the potential of mean force calculations\nusing explicit solvent (ES) models often fail to capture the solute\nstructure and assembly dynamics under various conditions. This study\nexamines the relationships between the parameter space of IS models\nand their predictive capabilities regarding solute structure and assembly\ndynamics, focusing on NaCl solutions and protein-bound silica colloidal\nnanoparticles. For NaCl solutions, models developed using concentration-dependent\ndielectric constants generally lack efficacy, and their effectiveness\nin predicting high-concentration liquid structures varies based on\nthe ES model used. Adjusting the dielectric constant value or the\nshape of the short-range potentials can improve the accuracy of these\nmodels in predicting liquid structures. For protein-bound silica colloidal\nnanoparticles, we find that interaction site size and model resolution\nsignificantly influence the prediction of nanoparticle assembly dynamics,\nwith different resolutions producing aggregates with distinct structures\nand connection modes. This study elucidates the complex relationship\nbetween parameter space in IS models and liquid structure and assembly\nkinetics, providing crucial insights for developing more accurate\nand predictive models.