Prediction And Analysis Of Small RNA During Zebrafish Early Developmental Stages

Small non-coding RNAs (sRNAs) of about20-30nucleotides (nt) play an essential role in a variety of animal developmental processes, such as embryonic, neuronal, muscle, and germline development. MicroRNAs (miRNAs) and Piwi-interacting RNAs (piRNAs), which are different in biogenesis and biological function, are two predominant types of sRNAs. miRNA (21-25) are bound by Ago-subfamily proteins of Argonaute (Ago)-family proteins, and function as post-transcriptional regulators by either translational repression, RNA degradation or both through an RNA-induced silencing complex (RISC). piRNAs (24-31) are associated with Piwi-subfamily proteins of Argonaute (Ago)-family proteins, and mainly necessary for germ cell maintenance and genome protection by silencing transposable elements.Rapid progress in NGS technologies has provided a great opportunity to investigate the sRNA transcriptome at an unprecedented sensitivity. However, it’s still a great challenge to analyze the deep sequencing data in an accurate and fast manner. Prediction of potentially new miRNAs and piRNA from highly heterogeneous short reads is particularly difficult and intriguing. Although a number of efforts have been contributed to this area, no tools were implemented specifically for analyzing zebrafish sRNA-seq data.Based on known zebrafish pre-miRNAs, we designed a zebrafish-specific algorithm of ZmirP (Zebrafish miRNA prediction) for pre-miRNAs prediction, with8new and57previously reported sequence and structure features. The performance and robustness of ZmirP were extensively evaluated by the leave-one-out (LOO) validation and n-fold cross-validations. By comparisons on zebrafish-specific dataset, ZmirP exhibits greater sensitivity of95.64%and specificity of98.84%, which is proved to be better than other existing approaches through comparison. Also, the performance of ZmirP is comparative with other tools for predicting human pre-miRNAs.The computational approaches for piRNA can classified into two major categories: piRNA cluster detection and piRNA transcript detention. Until now, proTRAC and piRNApredictor are only one software for detecting piRNA cluster and transcript, respectively. First, we searched the scientific literature from PubMed with the keyword of "piRNA" and "gsRNA" to obtain the experimentally verified piRNAs. Then a new algorithm of pirP (piRNA prediction) was developed for piRNA transcript prediction, with features of position-specific scoring and Z-curve. The performance and robustness of pirP were extensively evaluated by the leave-one-out validation and n-fold cross-validations. pirP exhibits greater sensitivity and specificity than piRNApredictor based on the experimentally verified data.Then we developed a specific platform as CSZ (Characterization of small RNAome for zebrafish) for the analysis of the high-through sequencing data. First, total reads were mapped to reference genome, while mapped reads were successively mapped to miRBase, Rfam, repeat, RefSeq mRNAs, and piRNABank to identify miRNAs, ncRNAs (including rRNA, tRNA, and snRNA/snoRNA), repeats, mRNAs and piRNAs. Many types of RNAs were annotated in repetitive sequences, such as rRNA, tRNA, snRNA/snoRNA, and repeat-associated piRNAs. If these annotations were not considered, the identification of the five types of sRNAs would be greatly underestimated. Thus, we designed a more efficient platform of CSZ, which first recalled rRNA, tRNA, and snRNA/snoRNA back to their own groups and then rescued piRNAs from the remaining repeats. For the unclassified reads, MIREAP and miRDeep2were used for the prediction of novel miRNAs. Because too many putative results were generated by MIREAP and miRDeep2, we adopted the ZmirP algorithm for further filtering potentially false positive hits.During early vertebrate development, various sRNAs, such as miRNAs and piRNAs are dynamically expressed for orchestrating the maternal-to-zygotic transition (MZT). Systematic analysis of expression profiles of zebrafish small RNAome will be greatly helpful for understanding the sRNA regulation during embryonic development. We first determined the expression profiles of sRNAs during eight distinct stages of early zebrafish development by sRNA-seq technology. Integrative analyses with a new computational platform of CSZ demonstrated an sRNA class transition from piRNAs to miRNAs as development proceeds. We observed that both the abundance and diversity of miRNAs are gradually increased, while the abundance is enhanced more dramatically than the diversity during development. However, although both the abundance and diversity of piRNAs are gradually decreased, the diversity was firstly increased then rapidly decreased. To evaluate the computational accuracy, the expression levels of four known miRNAs were experimentally validated. We also predicted25potentially novel miRNAs, whereas two candidates were verified by Northern blots.Taken together, our analyses revealed the piRNA to miRNA transition as a conserved mechanism in zebrafish, although two different types of sRNAs exhibit distinct expression dynamics in abundance and diversity, respectively. Our study not only generated a better understanding for sRNA regulations in early zebrafish development, but also provided a useful platform for analyzing sRNA-seq data.