TILD: A Strategy to Identify Cancer-related Genes Using Title Information in Literature Data

Abstract

After genome project in 1990s, researches which are involved with gene have been progressed. These studies unearthed that gene is cause of disease, and relations between gene and disease are important. In this reason, we proposed a strategy called TILD that identifies cancer-related genes using title information in literature data. To implement our method, we selected cancer-specific literature data from the online database. We then extracted genes using text mining. In the next step, we classified into two kinds for extracted genes using title information. If genes are located in title, then they are classified as hub genes. In the contrast, if genes are located in body, then they are classified as sub genes which are connected with hub genes. We iterated the processes for each paper to construct the cancer-specific local gene network. In the last step, we constructed global cancer-specific gene network by integrating all local gene network, and calculated a score for each gene based on analysis of the global gene network. We assumed that genes in title have meaningful relations with cancer, and other genes in the body are related with the title genes. For validation, we compared with other methods for the top 20 genes inferred by each approach. Our approach found more cancer-related genes than comparable methods.