Network Analysis of Biomedical Data Using REACH Natural Language Processing

Abstract

One of the core features of scientific work is reading and researching literature. Facing an unknown disease, beginning a new research project, or writing a paper to present your own research data are all good examples of why someone would run a wide literature search. However, it is difficult to stay up to date with the cutting edge of scientific research. A more recent approach that exceeds the limits of traditional manual research lies in automated literature analysis using natural language processing (NLP), something which is particularly relevant in complex biomedical research. NLP permits more rapid access to the information contained in scientific databases and may help to drastically increase the reproducibility of literature searches, allowing researchers to process all documents for a definite result. Network analysis techniques can be used to analyze the information extracted from literature, which often comes in the form of relationships between various biomedical entities. We used the REACH NLP tool to read 1.2 million papers in PubMed to extract mentions of biologically-relevant molecules and their relationship/interaction with other molecules. We then created an application - Visualizing Entities and Relationships in Text (VERIT), that allows users to generate network visualizations by querying the database we generated, which contains relationships between chemicals, proteins, genes, and phenotypes that were extracted by running REACH.