A Knowledge Graph Approach Towards Hidden Patterns Discovery From Biomedical Publications

Abstract: Biomedical research publications oen include significant scientific advancements in the biomedical domain. These publications can be interlinked in dierent aspects, such as having common keywords, covering a similar theme/topic, published by the same author(s), funded by a common funding agency, or belonging to a commonproject. Further, the visibility of links between dierent publications can be significantly useful to researchers, e.g., the keywords that interlink dierent authors can be useful to researchers for finding relevant biomedical researchers with common interests. However, it is diicult to interlink these biomedical publications since their bibliographic information is stored and accessible in fragments in research repositories. Hence, there is a need to investigate on how to interlink biomedical publications to uncover hidden patterns in order to achieve transparency in research. This study, following the design science methodology, investigates a knowledge graph based approach to interlink biomedical publications for uncovering hidden patterns. The study focuses on a use-case of biomedical publications by Örebro University between the years 1973-2021, which are 16626 in total. Biomedical concepts, authors and ailiation details, projects details, keywords, titles, and funding agencies details are extracted and conceptually modelled into a knowledge graph, which is later implemented in a graph database,i.e., Neo4j. Through demonstration of dierent queries on the database, this study finds that the implemented artefact enables greater visibility of links between publications, which in turn leads to visibility of hidden patterns between publications, authors, biomedical entities, funding organization, and projects. Furthermore, the artefact in this study presents information visually to users, which makes the results more transparent and easy to grasp. The artefact can be useful to both researchers and decision makers at an institute or funding agency, e.g., researchers can find other researchers or a potential funding agency based on a common interest, whereas decision makers will be able to get information about authors and funding details that are interlinked.