Computational prediction and characterization of long and circular non-coding RNAs and analysis of their non-canonical translation Seo-Won Choi

Abstract

Computational prediction and characterization of long and circular non-coding RNAs and analysis of their non-canonical translation Seo-Won Choi Department of Life Science The Graduate School of Hanyang University Non-coding RNAs are RNA molecules that do not code for protein but play various regulatory roles in cell. Besides the well-known non-coding RNAs such as rRNA and tRNA, the astounding development of RNA-seq and transcriptome assembly have brought a number of long non-coding RNAs (lncRNAs) and circular RNAs (circRNAs) to light. Currently, many of them are known to participate in stress- and cancer-related response. However, their vast number makes in challenging to characterize them. With the advancement in Ribo-seq, researchers have found that several lncRNAs harbored short ORFs and were associated with ribosomes, which makes their coding potential questionable. As most lncRNAs are discovered by transcriptome assembly, it is crucial to correctly discriminate them from novel protein-coding genes and erroneous assembly products. CircRNAs are covalently closed RNA molecules produced by back-splicing, which connects downstream splicing donor to upstream splicing acceptor. Although they are forced to underdo cap-independent translation, studies have found a small population of translated endogenous circRNAs. In addition, appropriate selection of internal ribosome entry site (IRES) and vector design has highlighted circular mRNAs as new mRNA vaccine, with the added advantage of their remarkable stability. Consequently, the efficiency at which circRNA is produced and translated has become the major issue of circRNA synthesis. However, the investigation of features that manage circularization efficiency is still at its early stage, with no known computational tools to offer such information. These findings point out that there are still uncovered parts in characterization of lncRNA and circRNA and the significance of devising computational tools that enable such investigation. In this paper, I suggested a novel computational tool that utilizes biological high-throughput data to accurately predict lncRNAs, named TERIUS. Furthermore, I identified several cis-and trans-factors that may regulate circularization efficiency and, based on the features, built a prediction model that gives probability of high-/low-circularization efficiency named POLARIS. Keywords: Long non-coding RNA, Circular RNA, Coding potential, Ribo-seq, Computational prediction