Field-SEA: a model for computing the solvation free energies of nonpolar, polar, and charged solutes in water

Abstract

Previous work describes a computational solvation model called semi-explicit assembly (SEA). The SEA water model computes the free energies of solvation of nonpolar and polar solutes in water with good efficiency and accuracy. However, SEA gives systematic errors in the solvation free energies of ions and charged solutes. Here, we describe field-SEA, an improved treatment that gives accurate solvation free energies of charged solutes, including monatomic and polyatomic ions and model dipeptides, as well as nonpolar and polar molecules. Field-SEA is computationally inexpensive for a given solute because explicit-solvent model simulations are relegated to a precomputation step and because it represents solvating waters in terms of a solute's free-energy field. In essence, field-SEA approximates the physics of explicit-model simulations within a computationally efficient framework. A key finding is that an atom's solvation shell inherits characteristics of a neighboring atom, especially strongly charged neighbors. Field-SEA may be useful where there is a need for solvation free-energy computations that are faster than explicit-solvent simulations and more accurate than traditional implicit-solvent simulations for a wide range of solutes.

Figure 1

A solvent–water molecule around a solute molecule. On the left, each solute atom (weakly charged) has a predefined radius, irrespective of its neighboring solute atoms, leading to the locus of water centers shown by the black dashed curve. On the right, one solute atom is strongly charged, leading to two consequences: its own solvating water molecule is pulled in tightly, and neighboring solute atoms have tighter water interactions too. We refer to the latter effect as an adaptive boundary. The energetic consequences can be large. Such effects may not be captured in simplified solvation models that treat atoms as having fixed radii, independent of neighboring atoms.

Figure 2

ΔG_solv as the function of a model LJ sphere (σ = 0.22 nm, ε = 0.06538 kJ/mol) charge for TIP3P, LPBE, and field-SEA. For comparison at infinite-dilution conditions, an Ewald correction is applied to the TIP3P and field-SEA results.

Figure 3

Field-SEA ΔG_solv for model diatomic solutes (triangles: Fixed r_w; circles: adaptive boundary) compared to TIP3P simulations. The line indicates the idealization of zero error.

Figure 4

Maps of the water–oxygen density around (A) weakly charged and (B) strongly charged diatomic solutes. The blue contours indicate water density greater than the bulk value, with the darker blue regions indicating the enhanced water probability density. The black line shows the nonadaptive fixed r_w boundary. For the weakly charged diatomic, the adaptive and nonadaptive boundaries coincide. The white line shows the adaptive boundary. For the charged diatomic, the adaptive and nonadaptive boundaries differ.

Figure 5

MD simulations of ΔG_solv for 504 neutral solutes (white), 35 molecular ions (orange), and 22 capped amino acid dipeptides (cyan) in TIP3P water, compared to (A) LPBE and (B) field-SEA.