Cosolvent interactions with biomolecules: Relating computer simulation data to experimental thermodynamic data

A major difficulty often encountered when the effects of cosolvents on the structure of biomolecules is studied by computer simulation is an inability to relate the simulation results to experimental thermodynamic data. In the absence of such a link, one cannot even determine the quality of the force field being used. Here we describe how quantitative thermodynamic data can be extracted from a computer simulation of a biomolecule in a cosolvent solution, and exactly how this is related to the corresponding experimental data. The approach involves a combination of the concept of preferential interactions and the use of Kirkwood-Buff theory to evaluate these interactions from simulation data. In particular, special attention is focused on the approximations made, the choice of activity scale for the biomolecule and cosolvent, and the use of the indistinguishable ion approach for the analysis of salt effects. The appropriate experimental thermodynamic data (volume fractions and activity derivatives) are provided for aqueous solutions of urea, guanidinium chloride, sodium chloride and 2,2,2,-trifluoroethanol. It is suggested that a determination of the simulated preferential interaction of a cosolvent with the native state of a protein under denaturing conditions provides the simplest test of available force fields, as it avoids simulations of the denatured state which are typically inaccessible with current computer power.