A novel evaluation measure for identifying drug targets from the biomedical literature

Identification of candidate target genes related to a particular disease is an important stage in drug development. A number of studies have extracted disease-related genes from the biomedical literature. We herein present a novel evaluation measure that identifies disease-associated genes and prioritizes the identified genes as drug target genes in terms of fewer side-effects using the biomedical literature. The proposed measure evaluates the specificity of a gene to a particular disease based on the number of diseases associated with the gene. The specificity of a gene is measured by means of, for example, term frequency-inverse document frequency (tf-idf), which is widely used in Web information retrieval. We assume that if a gene is chosen as a target gene for a disease, then side-effects are more likely to occur as the number of diseases associated with the gene increases. We verified the obtained ranking results by checking the ranks of known drug targets. As a result, 177 known drug targets were found to be ranked within the top 100 genes, and 21 drug targets were top ranked. The results suggest that the proposed measure is useful as a primary filter for extracting candidate target genes from a large number of genes.