Extraction of the pharmacological treatment features in literature-based discovery by applying cluster analysis

Literature-based discovery (LBD) facilitates the extraction of hidden relationships between a disease and chemical substances. As a methodology of LBD, we had previously proposed to apply cluster analysis to analyze the intermediate concepts between them. In this study, we compared the ranks of chemical substances predicted by our methodology to the original rank so as to validate possibilities for extraction of the heterogeneity in the relationships. As a result, we obtained two clusters. The rank for one cluster was similar to the original rank (ρ=0.956), but another was different (ρ=0.580). The different features from the original rank were obtained on Pharmacological Action Terms of MeSH terms.