Study On The Deciphering Unwinding Degree Of In Vivo MRNA Structure And Elaborating The Database Construction

As an ancient and essential biological macromolecule,RNA can be folded into the RNA structure,which can participate in complex cell activities.With the development of technology,a variety of prediction algorithms and experimental techniques have been used for RNA structure analysis.Among them,the most revolutionary is to combine high-throughput sequencing technology with chemical probes to detect single-stranded sites of wholetranscriptome RNA in vivo.It obtains the in vivo RNA structure of thousands of transcripts simultaneously in a single sample.Another technique,ribosome profiling,combines highthroughput sequencing to the ribosome footprint for obtaining the location of ribosomes in vivo with single-nucleotide accuracy.The emergence of these two techniques makes it possible to analyze the effects of intracellular ribosomes on mRNA structure.In this study,RNA structure data in Saccharomyces cerevisiae were first compared in vitro and in vivo.We adopt the Gini coefficient to evaluate the strength of the RNA structure.It concluded that the in vitro RNA structure of transcriptome is generally more robust than that in vivo and RNA structure in the untranslated region(UTR)is generally stronger than that in the coding region.In order to measure the degree of RNA unwinding in vivo,we proposed an indicator,"disappear structural trend(DIS)." We adopt five RNA structural features for each RNA structure through data processing,including ribosome density(RD),translation initiation ribosome density(INI),the relative position of the RNA structure(POS),the minimum free energy of RNA structure(MFE)and GC content(GC).For the correlation analysis of these five features and DIS,the positive correlation between ribosome density and DIS was the strongest,and it could be preliminarily concluded that the increase of ribosome density would lead to the RNA structural unwinding in vivo.Then,RNA structures were classified into the high degree of unwinding structure(HUS)and low degree of unwinding structure(LUS)according to DIS value,and a dichotomy model was established by using deep learning.Finally,a generalization and high precision deep learning model,Deep DRU,was obtained by the method of ten-fold cross-validation.The contributions of each structural features to the degree of RNA structural unwinding in vivo were resolved by the method of single-factor gradient adjustment,and the influencing factors were calculated.It's found that the translation initiation ribosome density contributed the most to the RNA structural unwinding.The next is ribosome density,and high ribosome density could cause a higher degree of mRNA structural unwinding because of ribosome occupation;At the same time,although the 5' end of mRNA has higher ribosome density,the RNA structural unwinding degree is lower.The study also found that the sequence features of the RNA structure,including the MFE and GC,contributed the least to the RNA structure in vivo.We also obtained 787 RNA regions with a stable high degree of structural unwinding(s HUS)and 1,905 RNA regions with a stable low degree of structural unwinding(s LUS)in Saccharomyces cerevisiae by model prediction.Finally,based on statistics,organize and implement visualization of the collected transcriptome RNA structure detection data,we established the transcriptome RNA structure database(RSVdb,https://taolab.nwsuaf.edu.cn/rsvdb).RSVdb currently contains 8 species,178 samples,and 626,225 transcripts with available DMS reactivity data.The database not only reprocesses and analyzes the RNA-structure-related sequencing data and generates interactive statistical charts and reports,but also supports the search of any RNA sequence,the online prediction with DMS reactivity restriction and interactive display of RNA structure.At the same time,the database contains detailed user manuals and provides download service of RNA structure data.To sum up,this research main through a combination of two kinds of HTS data and applying the technology of deep learning for modeling the degree of in vivo RNA structural unwinding.It expounds on the influence of the ribosome and other structural features on the mRNA structure.Finally,establish the transcriptome RNA structure database through the technology of server and network.In this study,we attempt to solve the biological problems by the fusion of multi-HTS-data and new technology,deepened the understanding of the ribosome affecting the RNA structure in vivo,and provided database service for future research on the regulation of RNA structure in the transcriptome.