A Computational Framework to Infer Prostate Cancer-Associated Long Noncoding RNAs and Analyses for Identifying a Competing Endogenous RNA Network.

Abstract

Background: Prostate cancer (PC) is the second leading cause of cancer death after lung cancer in men. Current biomarkers are ineffective for the treatment and management of the disease. Long noncoding RNAs (lncRNAs) are a heterogeneous group of transcripts that are involved in complex gene expression regulatory networks. Although lncRNAs have been suggested to be promising as future biomarkers, the connection between the majority of lncRNAs and human disease remains to be elucidated. One approach to elucidate the roles of lncRNAs in disease is through the development of computational models. For example, a novel computational model termed HyperGeometric distribution for LncRNA-Disease Association (HGLDA) has been developed. Such models need to be developed on a tumor-specific basis to better suit the particular problem. Methods: In this study, we constructed a potential pipeline through two models, HGLDA and pathway-based using data from several databases. To validate the obtained data, the expression levels of selected lncRNAs were investigated quantitatively in the DU-145, LNCaP, and PC3 PC cell lines using quantitative real-time PCR. Results: We obtained a number of lncRNAs from both models, many of which were filtered through several databases that ultimately resulted in identification of six high-value lncRNA targets. Their expression was correlated with one important component of the PI3K pathway, known to be related to PC. Conclusion: Through the assembly of a lncRNA-miRNAs-mRNA competing endogenous RNA network, we successfully predicted lncRNAs interfering with miRNAs and coding genes related to PC.