In line with the recent development of fragment-based drug design, a new computational method for mapping of small ligand molecules on protein surfaces is proposed. The method uses three-dimensional (3D) spatial distribution functions of the atomic sites of the ligand calculated using the molecular theory of solvation, known as the 3D reference interaction site model (3D-RISM) theory, to identify the most probable binding modes of ligand molecules. The 3D-RISM-based method is applied to the binding of several small organic molecules to thermolysin, in order to show its efficiency and accuracy in detecting binding sites. The results demonstrate that our method can reproduce the major binding modes found by X-ray crystallographic studies with sufficient precision. Moreover, the method can successfully identify some binding modes associated with a known inhibitor, which could not be detected by X-ray analysis. The dependence of ligand-binding modes on the ligand concentration, which essentially cannot be treated with other existing computational methods, is also investigated. The results indicate that some binding modes are readily affected by the ligand concentration, whereas others are not significantly altered. In the former case, it is the subtle balance in the binding affinity between the ligand and water that determines the dominant ligand-binding mode.