An alternative to manual literature review: a package for exploring biomedical concept graphs
Leslie Myint Macalester College
Abstract
Within the biomedical sciences, both contextual sensemaking of results and hypothesis generation require strong knowledge of how many different concepts and physical entities are connected. Traditionally, these endeavors have required investigators to manually mine the vast amounts of information contained in research papers. The sheer volume of accumulated knowledge poses problems for both the breadth and depth of literature review. One goal of biomedical text processing efforts is to facilitate this process through computation. The Semantic MEDLINE database, maintained by the National Library of Medicine, is a freely available resource that contains a rich store of biomedical relational information. In particular the database contains over 100 million subject-predicate-object relationships (called predications), that form a graph of relationships between biomedical concepts. The newly developed Bioconductor package rsemmed provides a set of tools for flexibly exploring mechanisms and conceptual connections within this graph. This package should be useful to anyone looking to computationally explore the literature behind biomedical ideas.
Keywords: literature-based discovery,biomedical knowledge graph