Integration of Annotated Phenotype, Gene and Chemical Text Data to Advance Exposome Informatics

Abstract

The field of phenomics has a range of biomedical informatics tools such as the Human Phenotype Ontology, providing a structured vocabulary with relationships between abnormal phenotype terms. Artificial intelligence has been widely used for entity extraction and tagging large corpora of text from PubMed and is reflected in applications such as PheneBank and PubTator. Phexpo is a tool for predicting chemical-phenotype relationships and vice-versa, although lacks the ability to decipher known relationships from unknown. Integration of these three resources can provide new meaningful relationships between phenotypes, genes and chemicals and has yet to be fully leveraged. Here we present a methodology to construct two new datasets for phenotype-gene and phenotype-chemical relationships and showcase how these datasets can be used to enhance exposome informatics.