Preferential solvation is a fundamental parameter for the interpretation of solubility and solute structural stability. The molecular basis for solute-solvent interactions can be obtained through distribution functions, and the thermodynamic connection to experimental data depends on the computation of distribution integrals, specifically Kirkwood-Buff integrals for the determination of preferential interactions. Standard radial distribution functions, however, are not convenient for the study of the solvation of complex, nonspherical solutes, as proteins. Here we show that minimum-distance distribution functions can be used to compute KB integrals while at the same time providing an insightful view of solute-solvent interactions at the molecular level. We compute preferential solvation parameters for Ribonuclease T1 in aqueous solutions of urea and trimethylamine N-oxide (TMAO) and show that, while macroscopic solvation shows that urea is preferentially bound to the protein surface and TMAO is preferentially excluded, both display specific density augmentations at the protein surface in dilute solutions. Therefore, direct protein-osmolyte interactions can play a role in the stability and activity of the protein even for preferentially hydrated systems. The generality of the distribution function and its natural connection to thermodynamic data suggest that it will be useful in general for the study of solvation in mixtures of structurally complex solutes and solvents.