GenDrux: A biomedical literature search system to identify gene expression-based drug sensitivity in breast cancer

Abstract

BackgroundThis paper describes the development of a web-based tool, GenDrux, which extracts and presents (over the Internet) information related to the disease-gene-drug nexus. This information is archived from the relevant biomedical literature using automated methods. GenDrux is designed to alleviate the difficulties of manually processing the vast biomedical literature to identify disease-gene-drug relationships. GenDrux will evolve with the literature without additional algorithmic modifications.ResultsGenDrux, a pilot system, is developed in the domain of breast cancer and can be accessed at http://www.microarray.uab.edu/drug_gene.pl. GenDrux can be queried based on drug, gene and/or disease name. From over 8,000 relevant abstracts from the biomedical literature related to breast cancer, we have archived a corpus of more than 4,000 articles that depict gene expression-drug activity relationships for breast cancer and related cancers. The archiving process has been automated.ConclusionsThe successful development, implementation, and evaluation of this and similar systems when created may provide clinicians with a tool for literature management, clinical decision making, thus setting the platform for personalized therapy in the future.

Automated knowledge extraction from polymer literature using natural language processing

“…An application is marked as ''corresponding'' to a polymer CNE if it occurs in a context window of size 4 around it. Similar approaches have been followed in literature to infer relationships between entities from co-occurrence (Crasto et al, 2011;Li et al (2009)). We validate our approach by randomly selecting polymer-application pairs that appear at least 5 times in our corpus and verify that the corresponding sentence in the paper where the polymer and application co-occur is indeed a true relationship and not a random co-occurrence.…”

Information Retrieval and Text Mining Technologies for Chemistry

“…The simplest approach for RE, a sort of baseline method, relies on comention, cooccurrence, or cocitation of entities within a specific context, defined as individual sentences, abstracts, paragraphs, or whole documents. , Simple co-occurrence, without any additional constraints, does represent the overall upper boundary in terms of relation extraction recall and lower boundary in terms of precision. Relation extraction using co-occurrence assumes that if entities are mentioned together in a particular unit of text, they should have some sort of association.…”

An eUtils toolset and its use for creating a pipeline to link genomics and proteomics analyses to domain-specific biomedical literature

“…Literature Surveillance and filtering of results to areas of specific research interest. GenDrux[ 11 ], queries PubMed to identify disease-gene-drug relationships; the pilot implementation has been applied to the breast cancer area. The present paper’s work falls into this category.…”