PEA-m6A: an ensemble learning framework for accurately predicting N6-methyladenosine modifications in plants.

N 6-methyladenosine (m6A), which is the mostly prevalent modification in eukaryotic mRNAs, is involved in gene expression regulation and many RNA metabolism processes. Accurate prediction of m6A modification is important for understanding its molecular mechanisms in different biological contexts. However, most existing models have limited range of application and are species-centric. Here we present PEA-m6A, a unified, modularized and parameterized framework that can streamline m6A-Seq data analysis for predicting m6A-modified regions in plant genomes. The PEA-m6A framework builds ensemble learning-based m6A prediction models with statistic-based and deep learning-driven features, achieving superior performance with an improvement of 6.7% to 23.3% in the area under precision-recall curve compared with state-of-the-art regional-scale m6A predictor WeakRM in 12 plant species. Especially, PEA-m6A is capable of leveraging knowledge from pretrained models via transfer learning, representing an innovation in that it can improve prediction accuracy of m6A modifications under small-sample training tasks. PEA-m6A also has a strong capability for generalization, making it suitable for application in within- and cross-species m6A prediction. Overall, this study presents a promising m6A prediction tool, PEA-m6A, with outstanding performance in terms of its accuracy, flexibility, transferability, and generalization ability. PEA-m6A has been packaged using Galaxy and Docker technologies for ease of use and is publicly available at https://github.com/cma2015/PEA-m6A.

Metagenomics insights into responses of rhizobacteria and their alleviation role in licorice allelopathy.

BACKGROUND: Allelopathy is closely associated with rhizosphere biological processes, and rhizosphere microbial communities are essential for plant development. However, our understanding of rhizobacterial communities under influence of allelochemicals in licorice remains limited. In the present study, the responses and effects of rhizobacterial communities on licorice allelopathy were investigated using a combination of multi-omics sequencing and pot experiments, under allelochemical addition and rhizobacterial inoculation treatments. RESULTS: Here, we demonstrated that exogenous glycyrrhizin inhibits licorice development, and reshapes and enriches specific rhizobacteria and corresponding functions related to glycyrrhizin degradation. Moreover, the Novosphingobium genus accounted for a relatively high proportion of the enriched taxa and appeared in metagenomic assembly genomes. We further characterized the different capacities of single and synthetic inoculants to degrade glycyrrhizin and elucidated their distinct potency for alleviating licorice allelopathy. Notably, the single replenished N (Novosphingobium resinovorum) inoculant had the greatest allelopathy alleviation effects in licorice seedlings. CONCLUSIONS: Altogether, the findings highlight that exogenous glycyrrhizin simulates the allelopathic autotoxicity effects of licorice, and indigenous single rhizobacteria had greater effects than synthetic inoculants in protecting licorice growth from allelopathy. The results of the present study enhance our understanding of rhizobacterial community dynamics during licorice allelopathy, with potential implications for resolving continuous cropping obstacle in medicinal plant agriculture using rhizobacterial biofertilizers. Video Abstract.

Large-Scale Detection of Telomeric Motif Sequences in Genomic Data Using TelFinder.

Telomeres are regions of tandem repeated sequences at the ends of linear chromosomes that protect against DNA damage and chromosome fusion. Telomeres are associated with senescence and cancers and have attracted the attention of an increasing number of researchers. However, few telomeric motif sequences are known. Given the mounting interest in telomeres, an efficient computational tool for the de novo detection of the telomeric motif sequence of new species is needed since experimental-based methods are costly in terms of time and effort. Here, we report the development of TelFinder, an easy-to-use and freely available tool for the de novo detection of telomeric motif sequences from genomic data. The vast quantity of readily available genomic data makes it possible to apply this tool to any species of interest, which will undoubtedly inspire studies requiring telomeric repeat information and improve the utilization of these genomic data sets. We have tested TelFinder on telomeric sequences available in the Telomerase Database, and the detection accuracy reaches 90%. In addition, variation analyses in telomere sequences can be performed by TelFinder for the first time. The telomere variation preference of different chromosomes and even at the ends of the chromosome can provide clues regarding the underlying mechanisms of telomeres. Overall, these results shed new light on the divergent evolution of telomeres. IMPORTANCE Telomeres are reported to be highly correlated with the cell cycle and aging. As a result, research on telomere composition and evolution has become more and more urgent. However, using experimental methods to detect telomeric motif sequences is slow and costly. To combat this challenge, we developed TelFinder, a computational tool for the de novo detection of the telomere composition only using genomic data. In this study, we showed that a lot of complicated telomeric motifs could be identified by TelFinder only using genomic data. In addition, TelFinder can be used to check variation analyses in telomere sequences, which could lead to a deeper understanding of telomere sequences.

Natural selection pressure exerted on "Silent" mutations during the evolution of SARS-CoV-2: Evidence from codon usage and RNA structure.

From the first emergence of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) till now, multiple mutations that caused synonymous and nonsynonymous substitutions have accumulated. Among them, synonymous substitutions were regarded as "silent" mutations that received less attention than nonsynonymous substitutions that cause amino acid variations. However, the importance of synonymous substitutions can not be neglected. This research focuses on synonymous substitutions on SARS-CoV-2 and proves that synonymous substitutions were under purifying selection in its evolution. The evidence of purifying selection is provided by comparing the mutation number per site in coding and non-coding regions. We then study the two forces of purifying selection: synonymous codon usage and RNA secondary structure. Results show that the codon usage optimization leads to an adapted codon usage towards humans. Furthermore, our results show that the maintenance of RNA secondary structure causes the purifying of synonymous substitutions in the structural region. These results explain the selection pressure on synonymous substitutions during the evolution of SARS-CoV-2.

A detailed comparative analysis of codon usage bias in Alongshan virus.

Alongshan virus (ALSV) is an emerging tick-borne pathogen that infects humans, causing febrile disease. ALSV uses Ixodes Persulcatus ticks to infect humans with a wide range of signs, from asymptomatic to encephalitis-like syndrome. There is an increasing public health concern about the ALSV infection. To get insight into the impacts of viral relations with their hosts on viral ability, survival, and evasion from hosts immune systems remain unknown. The codon usage is a driving force in viral genome evolution; therefore, we enrolled 41 ALSV strains in codon usage analysis to elucidate the molecular evolutionary dynamics of ALSV. The results indicate that the overall codon usage among ALSV isolates is relatively similar and slightly biased. Base compositions for the cds were in order of G >A >C >U and in the third position of codons G3 >A3 >C3 >T3. The RSCU values revealed that the more frequently used codons were mostly GC ended. Different codon preferences in ALSV genes in relation to codon usage of H. sapiens and Ixodes Persulcatus genes were found. Neutrality plot was determined to reveal the superiority of natural selection over directional mutation pressure in causing CUB based on GC12 versus GC3 contents. The results of these studies suggest that the emergence of ALSV in China, Russia and Finland may also be reflected in ALSV codon usage. Altogether, the presence of both mutation pressure and natural selection effect in shaping the codon usage patterns of ALSV.

Genome-wide characterization, evolution, structure, and expression analysis of the F-box genes in Caenorhabditis.

BACKGROUND: F-box proteins represent a diverse class of adaptor proteins of the ubiquitin-proteasome system (UPS) that play critical roles in the cell cycle, signal transduction, and immune response by removing or modifying cellular regulators. Among closely related organisms of the Caenorhabditis genus, remarkable divergence in F-box gene copy numbers was caused by sizeable species-specific expansion and contraction. Although F-box gene number expansion plays a vital role in shaping genomic diversity, little is known about molecular evolutionary mechanisms responsible for substantial differences in gene number of F-box genes and their functional diversification in Caenorhabditis. Here, we performed a comprehensive evolution and underlying mechanism analysis of F-box genes in five species of Caenorhabditis genus, including C. brenneri, C. briggsae, C. elegans, C. japonica, and C. remanei. RESULTS: Herein, we identified and characterized 594, 192, 377, 39, 1426 F-box homologs encoding putative F-box proteins in the genome of C. brenneri, C. briggsae, C. elegans, C. japonica, and C. remanei, respectively. Our work suggested that extensive species-specific tandem duplication followed by a small amount of gene loss was the primary mechanism responsible for F-box gene number divergence in Caenorhabditis genus. After F-box gene duplication events occurred, multiple mechanisms have contributed to gene structure divergence, including exon/intron gain/loss, exonization/pseudoexonization, exon/intron boundaries alteration, exon splits, and intron elongation by tandem repeats. Based on high-throughput RNA sequencing data analysis, we proposed that F-box gene functions have diversified by sub-functionalization through highly divergent stage-specific expression patterns in Caenorhabditis species. CONCLUSIONS: Massive species-specific tandem duplications and occasional gene loss drove the rapid evolution of the F-box gene family in Caenorhabditis, leading to complex gene structural variation and diversified functions affecting growth and development within and among Caenorhabditis species. In summary, our findings outline the evolution of F-box genes in the Caenorhabditis genome and lay the foundation for future functional studies.

Edging on Mutational Bias, Induced Natural Selection From Host and Natural Reservoirs Predominates Codon Usage Evolution in Hantaan Virus.

The molecular evolutionary dynamics that shape hantaviruses' evolution are poorly understood even now, besides the contribution of virus-host interaction to their evolution remains an open question. Our study aimed to investigate these two aspects in Hantaan virus (HTNV)-the prototype of hantaviruses and an emerging zoonotic pathogen that infects humans, causing hemorrhagic fever with renal syndrome (HFRS): endemic in Far East Russia, China, and South Korea-via a comprehensive, phylogenetic-dependent codon usage analysis. We found that host- and natural reservoir-induced natural selection is the primary determinant of its biased codon choices, exceeding the mutational bias effect. The phylogenetic analysis of HTNV strains resulted in three distinct clades: South Korean, Russian, and Chinese. An effective number of codon (ENC) analysis showed a slightly biased codon usage in HTNV genomes. Nucleotide composition and RSCU analyses revealed a significant bias toward A/U nucleotides and A/U-ended codons, indicating the potential influence of mutational bias on the codon usage patterns of HTNV. Via ENC-plot, Parity Rule 2 (PR2), and neutrality plot analyses, we would conclude the presence of both mutation pressure and natural selection effect in shaping the codon usage patterns of HTNV; however, natural selection is the dominant factor influencing its codon usage bias. Codon adaptation index (CAI), Relative codon deoptimization index (RCDI), and Similarity Index (SiD) analyses uncovered the intense selection pressure from the host (Human) and natural reservoirs (Striped field mouse and Chinese white-bellied rat) in shaping HTNV biased codon choices. Our study clearly revealed the evolutionary processes in HTNV and the role of virus-host interaction in its evolution. Moreover, it opens the door for a more comprehensive codon usage analysis for all hantaviruses species to determine their molecular evolutionary dynamics and adaptability to several hosts and environments. We believe that our research will help in a better and deep understanding of HTNV evolution that will serve its future basic research and aid live attenuated vaccines design.

Analysis of codon usage patterns and influencing factors in rice tungro bacilliform virus.

Rice tungro bacilliform virus (RTBV) belongs to genus Tungrovirus within the family Caulimoviridae harbors circular double-stranded DNA (dsDNA). Rice tungro disease (RTD) caused by RTBV, responsible for severe rice yield losses in South and Southeast Asia. Here, we performed a systematic evolutionary and codon usage bias (CUB) analysis of RTBV genome sequences. We analysed different bioinformatics techniques to calculate the nucleotide compositions, the relative synonymous codon usage (RSCU), and other indices. The results indicated slightly or low codon usage bias in RTBV isolates. Mutation and natural selection pressures have equally contributed to this low codon usage bias. Additionally, multiple factors such as host, geographical distribution also affect codon usage patterns in RTBV genomes. RSCU analysis revealed that RTBV shows mutation bias and prefers A and U ended codons to code amino acids. Codon usage patterns of RTBV were also found to be influenced by its host. This indicates that RTBV have evolved codon usage patterns that are specific to its host. The findings from this study are expected to increase our understanding of factors leading to viral evolution and fitness with respect to hosts and the environment.

Altered Metabolic Strategies: Elaborate Mechanisms Adopted by Oenococcus oeni in Response to Acid Stress.

Oenococcus oeni plays a key role in inducing malolactic fermentation in wine. Acid stress is often encountered under wine conditions. However, the lack of systematic studies of acid resistance mechanisms limits the downstream fermentation applications. In this study, the acid responses of O. oeni were investigated by combining transcriptome, metabolome, and genome-scale metabolic modeling approaches. Metabolite profiling highlighted the decreased abundance of nucleotides under acid stress. The gene-metabolite bipartite network showed negative correlations between nucleotides and genes involved in ribosome assembly, translation, and post-translational processes, suggesting that stringent response could be activated under acid stress. Genome-scale metabolic modeling revealed marked flux rerouting, including reallocation of pyruvate, attenuation of glycolysis, utilization of carbon sources other than glucose, and enhancement of nucleotide salvage and the arginine deiminase pathway. This study provided novel insights into the acid responses of O. oeni, which will be useful for designing strategies to address acid stress in wine malolactic fermentation.

RSVdb: a comprehensive database of transcriptome RNA structure.

RNA fulfills a crucial regulatory role in cells by folding into a complex RNA structure. To date, a chemical compound, dimethyl sulfate (DMS), has been developed to probe the RNA structure at the transcriptome level effectively. We proposed a database, RSVdb (https://taolab.nwafu.edu.cn/rsvdb/), for the browsing and visualization of transcriptome RNA structures. RSVdb, including 626 225 RNAs with validated DMS reactivity from 178 samples in eight species, supports four main functions: information retrieval, research overview, structure prediction and resource download. Users can search for species, studies, transcripts and genes of interest; browse the quality control of sequencing data and statistical charts of RNA structure information; preview and perform online prediction of RNA structures in silico and under DMS restraint of different experimental treatments and download RNA structure data for species and studies. Together, RSVdb provides a reference for RNA structure and will support future research on the function of RNA structure at the transcriptome level.

Origin and Evolution of Fusidane-Type Antibiotics Biosynthetic Pathway through Multiple Horizontal Gene Transfers.

Fusidane-type antibiotics represented by fusidic acid, helvolic acid, and cephalosporin P1 have very similar core structures, but they are produced by fungi belonging to different taxonomic groups. The origin and evolution of fusidane-type antibiotics biosynthetic gene clusters (BGCs) in different antibiotics producing strains remained an enigma. In this study, we investigated the distribution and evolution of the fusidane BGCs in 1,284 fungal genomes. We identified 12 helvolic acid BGCs, 4 fusidic acid BGCs, and 1 cephalosporin P1 BGC in Pezizomycotina fungi. Phylogenetic analyses indicated six horizontal gene transfer (HGT) events in the evolutionary trajectory of the BGCs, including 1) three transfers across Eurotiomycetes and Sordariomycetes classes; 2) one transfer between genera under Sordariomycetes class; and 3) two transfers within Aspergillus genus under Eurotiomycetes classes. Finally, we proposed that the ancestor of fusidane BGCs would be originated from the Zoopagomycota by ancient HGT events according to the phylogenetic trees of key enzymes in fusidane BGCs (OSC and P450 genes). Our results extensively clarify the evolutionary trajectory of fusidane BGCs by HGT among distantly related fungi and provide new insights into the evolutionary mechanisms of metabolic pathways in fungi.

CRISPR-CBEI: a Designing and Analyzing Tool Kit for Cytosine Base Editor-Mediated Gene Inactivation.

Base editing is a promising technique, allowing precise single-base mutagenesis in genomes without double-strand DNA breaks or donor templates. Cytosine base editors (CBEs) convert cytosine to thymidine. In particular, CBEs can transform four codons, CAA, CAG, CGA, and TGG, into stop codons, providing a new means to rapidly inactivate a gene of interest and enabling loss-of-function study in recombination-deficient species and the construction of gene-inactivation libraries. However, designing single guide RNAs (sgRNAs) for gene inactivation is more complicated and more restricted in applicability than using the lustered regularly interspaced short palindromic repeats (CRISPR)/CRISPR-associated (CRISPR/Cas9) system only, especially for researchers who do not specialize in the bioinformatics skills needed to design and evaluate sgRNAs. Here, we present a new user-friendly designing tool kit, namely, CRISPR-CBEI (cytosine base editor-mediated gene inactivation), including a Web tool and a command-line tool. The Web tool is dedicated to the design of sgRNAs for CBE-mediated gene inactivation and integrates various functions, including open reading frame (ORF) identification, CBE customization, sgRNA designing, summarizing, and front-end off-target searching against user-defined unlimited-file-size local genome files without the necessity of uploading to the server. The command-line version serves the same purpose but for a larger number of coding DNA sequences (CDSs), for instance, for designing a CBE-inactivation library in a target species which provides comprehensive evaluations of CBEs and target genomes. We envision that this tool would contribute to CBE-inactivation design.IMPORTANCE Life science has been in pursuit of precise and efficient genome editing in living cells since the very beginning of the first restriction cloning attempt. The introduction of RNA-guided CRISPR-associated (Cas) nucleases contributed to this ultimate goal through their ability to deliver a double-strand break (DSB) to a precise target location in various species, obsoleting the preceding editing tools, such as zinc-finger nucleases (ZFNs) and transcription activator-like effector nucleases (TALENs). The derivative technology, base editing, combines the catalytically inactivated Cas nuclease and nucleotide deaminase and mediates the genetic modifications at single-nucleotide precision without introducing a DSB. Moreover, the cytosine base editors (CBEs) are able to transform multiple codons into stop codons, rapidly inactivating a gene of interest and enabling loss-of-function study in some recombination-deficient species. Here, we present the CRISPR-CBEI tool kit to assist the design of sgRNAs for CBE-mediated gene inactivation.

Codon Usage Bias Analysis of Bluetongue Virus Causing Livestock Infection.

Bluetongue virus (BTV) is a double-stranded RNA virus with multiple segments and belongs to the genus Orbivirus within the family Reoviridae. BTV is spread to livestock through its dominant vector, biting midges of genus Culicoides. Although great progress has been made in genomic analyses, it is not fully understood how BTVs adapt to their hosts and evade the host's immune systems. In this study, we retrieved BTV genome sequences from the National Center for Biotechnology Information (NCBI) database and performed a comprehensive research to explore the codon usage patterns in 50 BTV strains. We used bioinformatic approaches to calculate the relative synonymous codon usage (RSCU), codon adaptation index (CAI), effective number of codons (ENC), and other indices. The results indicated that most of the overpreferred codons had A-endings, which revealed that mutational pressure was the major force shaping codon usage patterns in BTV. However, the influence of natural selection and geographical factors cannot be ignored on viral codon usage bias. Based on the RSCU values, we performed a comparative analysis between BTVs and their hosts, suggesting that BTVs were inclined to evolve their codon usage patterns that were comparable to those of their hosts. Such findings will be conducive to understanding the elements that contribute to viral evolution and adaptation to hosts.

Implication of the gut microbiome composition of type 2 diabetic patients from northern China.

Emerging evidence has suggested the association of the gut microbiome with some human diseases, including type 2 diabetes (T2D). In this study, we analyzed the gut microbiota from a cohort of healthy and diabetic Chinese individuals from Northern China. Pyrosequencing of the V4V5 region of 16S rRNA genes revealed a significant decrease in the gut microbiota diversity of diabetic patients as compared to healthy individuals. Butyrate-producing bacteria such as Bifidobacterium and Akkermansia were significantly decreased in diabetic patients. Furthermore, the abundance of Dorea was significantly increased in T2D individuals and negatively correlated with the abundance of butyrate-producing bacteria. The increase of Dorea could play a role in the development of T2D and has been previously overlooked. Importantly, functional analysis of the gut microbiome revealed for the first time that increased levels of butyrate production via transferases and the degradation of several amino acids due to gut microbial metabolism have strong correlations with T2D in Northern China. Moreover, the potential of gut microbiota-based classifiers to identify individuals with a high risk for T2D has been demonstrated in this study. Taken together, our findings have revealed a previously unappreciated association of the gut microbiome with T2D and have also suggested that changes in gut microbiota may be used to identify individuals at high risk for T2D.

Vasa deferentia and associated structures of the male Panorpodes kuandianensis (Mecoptera: Panorpodidae).

The male reproductive system may provide significant evidence for the taxonomic and phylogenetic analyses of insects. However, current knowledge of the male reproductive system in Mecoptera is mainly concentrated on the external genitalia, and is rarely involved in the internal reproductive system. Here, we investigated the morphology and the fine structure of the vasa deferentia and associated structures of the male reproductive system of Panorpodes kuandianensis Zhong et al., 2011 (Panorpodidae) using light, scanning, and transmission electron microscopy. The male reproductive system of P. kuandianensis consists of a pair of symmetrical testes with three tubular testicular follicles, two epididymides, two distinctly partitioned vasa deferentia, a pair of mesadenia, one ejaculatory sac, and the external genitalia. A pair of expanded seminal vesicles are modified from the median part of the vasa deferentia, and evolve into secretory organs. The seminal vesicles have elongated cylindrical epithelial cells, which contain abundant secretory materials in the cytoplasm and form a small central lumen, likely serving a secretory function rather than provisionally storing sperm as in most other insects. Alternatively, the sperm are stored temporarily in the epididymis, the greatly coiled portion of the vasa deferentia. The morphology of the male reproductive system supports the close relationships of Panorpidae and Panorpodidae.

Genomic insight into the origins and evolution of symbiosis genes in Phaseolus vulgaris microsymbionts.

BACKGROUND: Phaseolus vulgaris (common bean) microsymbionts belonging to the bacterial genera Rhizobium, Bradyrhizobium, and Ensifer (Sinorhizobium) have been isolated across the globe. Individual symbiosis genes (e.g., nodC) of these rhizobia can be different within each genus and among distinct genera. Little information is available about the symbiotic structure of indigenous Rhizobium strains nodulating introduced bean plants or the emergence of a symbiotic ability to associate with bean plants in Bradyrhizobium and Ensifer strains. Here, we sequenced the genomes of 29 representative bean microsymbionts (21 Rhizobium, four Ensifer, and four Bradyrhizobium) and compared them with closely related reference strains to estimate the origins of symbiosis genes among these Chinese bean microsymbionts. RESULTS: Comparative genomics demonstrated horizontal gene transfer exclusively at the plasmid level, leading to expanded diversity of bean-nodulating Rhizobium strains. Analysis of vertically transferred genes uncovered 191 (out of the 2654) single-copy core genes with phylogenies strictly consistent with the taxonomic status of bacterial species, but none were found on symbiosis plasmids. A common symbiotic region was wholly conserved within the Rhizobium genus yet different from those of the other two genera. A single strain of Ensifer and two Bradyrhizobium strains shared similar gene content with soybean microsymbionts in both chromosomes and symbiotic regions. CONCLUSIONS: The 19 native bean Rhizobium microsymbionts were assigned to four defined species and six putative novel species. The symbiosis genes of R. phaseoli, R. sophoriradicis, and R. esperanzae strains that originated from Mexican bean-nodulating strains were possibly introduced alongside bean seeds. R. anhuiense strains displayed distinct host ranges, indicating transition into bean microsymbionts. Among the six putative novel species exclusive to China, horizontal transfer of symbiosis genes suggested symbiosis with other indigenous legumes and loss of originally symbiotic regions or non-symbionts before the introduction of common bean into China. Genome data for Ensifer and Bradyrhizobium strains indicated symbiotic compatibility between microsymbionts of common bean and other hosts such as soybean.

Exploring the evolutionary dynamics of Rhizobium plasmids through bipartite network analysis.

The genus Rhizobium usually has a multipartite genome architecture with a chromosome and several plasmids, making these bacteria a perfect candidate for plasmid biology studies. As there are no universally shared genes among typical plasmids, network analyses can complement traditional phylogenetics in a broad-scale study of plasmid evolution. Here, we present an exhaustive analysis of 216 plasmids from 49 complete genomes of Rhizobium by constructing a bipartite network that consists of two classes of nodes, the plasmids and homologous protein families that connect them. Dissection of the network using a hierarchical clustering strategy reveals extensive variety, with 34 homologous plasmid clusters. Four large clusters including one cluster of symbiotic plasmids and two clusters of chromids carrying some truly essential genes are widely distributed among Rhizobium. In contrast, the other clusters are quite small and rare. Symbiotic clusters and rare accessory clusters are exogenetic and do not appear to have co-evolved with the common accessory clusters; the latter ones have a large coding potential and functional complementarity for different lifestyles in Rhizobium. The bipartite network also provides preliminary evidence of Rhizobium plasmid variation and formation including genetic exchange, plasmid fusion and fission, exogenetic plasmid transfer, host plant selection, and environmental adaptation.

Deciphering the rules of mRNA structure differentiation in Saccharomyces cerevisiae in vivo and in vitro with deep neural networks.

The structure of mRNA in vivo is unwound to some extent in response to multiple factors involved in the translation process, resulting in significant differences from the structure of the same mRNA in vitro. In this study, we have proposed a novel application of deep neural networks, named DeepDRU, to predict the degree of mRNA structure unwinding in vivo by fitting five quantifiable features that may affect mRNA folding: ribosome density (RD), minimum folding free energy (MFE), GC content, translation initiation ribosome density (INI) and mRNA structure position (POS). mRNA structures with adjustment of the simulated structural features were designed and then fed into the trained DeepDRU model. We found unique effect regions of these five features on mRNA structure in vivo. Strikingly, INI is the most critical factor affecting the structure of mRNA in vivo, and structural sequence features, including MFE and GC content, have relatively smaller effects. DeepDRU provides a new paradigm for predicting the unwinding capability of mRNA structure in vivo. This improved knowledge about the mechanisms of factors influencing the structural capability of mRNA to unwind will facilitate the design and functional analysis of mRNA structure in vivo.