ceRNAR: An R package for identification and analysis of ceRNA-miRNA triplets

Competitive endogenous RNA (ceRNA) represents a novel mechanism of gene regulation that controls several biological and pathological processes. Recently, an increasing number of in silico methods have been developed to accelerate the identification of such regulatory events. However, there is still a need for a tool supporting the hypothesis that ceRNA regulatory events only occur at specific miRNA expression levels. To this end, we present an R package, ceRNAR, which allows identification and analysis of ceRNA-miRNA triplets via integration of miRNA and RNA expression data. The ceRNAR package integrates three main steps: (i) identification of ceRNA pairs based on a rank-based correlation between pairs that considers the impact of miRNA and a running sum correlation statistic, (ii) sample clustering based on gene-gene correlation by circular binary segmentation, and (iii) peak merging to identify the most relevant sample patterns. In addition, ceRNAR also provides downstream analyses of identified ceRNA-miRNA triplets, including network analysis, functional annotation, survival analysis, external validation, and integration of different tools. The performance of our proposed approach was validated through simulation studies of different scenarios. Compared with several published tools, ceRNAR was able to identify true ceRNA triplets with high sensitivity, low false-positive rates, and acceptable running time. In real data applications, the ceRNAs common to two lung cancer datasets were identified in both datasets. The bridging miRNA for one of these, the ceRNA for MAP4K3, was identified by ceRNAR as hsa-let-7c-5p. Since similar cancer subtypes do share some biological patterns, these results demonstrated that our proposed algorithm was able to identify potential ceRNA targets in real patients. In summary, ceRNAR offers a novel algorithm and a comprehensive pipeline to identify and analyze ceRNA regulation. The package is implemented in R and is available on GitHub (https://github.com/ywhsiao/ceRNAR).Author summary: The gene expression regulating mechanisms in humans are complex as many regulators are highly connected and are compensatory to each other. Not to mention, a large proportion of the potential interactions between miRNA and gene expression remain unclear due to the challenges and difficulties of performing biological experiments to validate them. With the advancement in high-throughput genomic technologies, massive data of different molecules can be generated within a short period of time. However, utilizing such massive data towards unveiling the regulatory relationships through computational methods and statistical models poses a bottleneck. To address this issue, we present an R package, ceRNAR, that enables researchers to explore and identify potential competing endogeneous RNA (ceRNA) events through three consecutive steps, and provides novel biological insights into the analytical results. ceRNA constitutes of a set of different RNAs that compete with messenger RNA for interacting with a given miRNA, towards gene expression regulation. Through our proposed tool, users can avail a novel algorithm and a comprehensive pipeline for identifying novel regulators and interactions among miRNA and messenger RNA that may potentially explain biological and pathological processes.