Increasingly greater number of research papers are being published, which still serve as a main source of new findings. Despite the current availability of highly efficient search engines, ironically, it has become more difficult to efficiently collect and process newly reported data from vast amounts of papers at once. One important information that should be gathered from literature in a systematic manner is gene-protein-reaction (GPR) associations, which corresponds to an extended effort for genome annotation. Availability of GPR associations itself is a valuable information for studying genotype-phenotype associations of an organism, and can also be effectively used for development of computational models, for example genome-scale metabolic model. Here, we introduce a Python-based scientific text navigation system that efficiently collects information on GPR associations from the PubMed database. BioBERT was applied to retrieve information on five entities, including
species, genes, proteins, chemicals, and metabolites, and GPR associations were subsequently reconstructed. The scientific text navigation system developed in this study will allow more efficient and systematic collection of biological information from a large volume of literature.