We developed a new method for the classification of chemical compounds and protein pockets and applied it to a random screening experiment for macrophage migration inhibitory factor (MIF). The principal component analysis (PCA) method was applied to the protein-compound interaction matrix, which was given by thorough docking calculations between a set of many protein pockets and chemical compounds. Each compound and protein pocket was depicted as a point in the PCA spaces of compounds and proteins, respectively. This method was applied to distinguish active compounds from negative compounds of MIF. A random screening experiment for MIF was performed, and our method revealed that the active compounds were localized in the PCA space of compounds, while the negative compounds showed a wide distribution. Furthermore, protein pockets, which bind similar compounds, were classified and were found to form a cluster in the PCA space.