Promoter region of protein-coding genes are gradually being well understood, yet no comparable studies exist for the promoter of long non-coding RNA (lncRNA) genes which has emerged as a global potential regulator in multiple cellular process and different diseases for human. To understand the difference in the transcriptional regulation pattern of these genes, previously, we proposed a machine learning based model to classify the promoter of protein-coding genes and lncRNA genes. In this study, we are presenting DeepCNPP (deep coding non-coding promoter predictor), an improved model based on deep learning (DL) framework to classify the promoter of lncRNA genes and protein-coding genes. We used convolution neural network (CNN) based deep network to classify the promoter of these two broad categories of human genes. Our computational model, built upon the sequence information only, was able to classify these two groups of promoters from human at a rate of 83.34% accuracy and outperformed the existing model. Further analysis and interpretation of the output from DeepCNPP architecture will enable us to understand the difference in transcription regulatory pattern for these two groups of genes.

Indiana University-Purdue University Indianapolis (IUPUI) / Several experimental methods has been developed for the study of the central dogma since late 20th century. Protein mass spectrometry and next generation sequencing (including DNA-Seq and RNA-Seq) forms a triangle of experimental methods, corresponding to the three vertices of the central dogma, i.e., DNA, RNA and protein. Numerous RNA sequencing and protein mass spectrometry experiments has been carried out in attempt to understand how the expression change of known genes affect biological functions in various of organisms, however, it has been once overlooked that the result data of these experiments are in fact holograms which also reveals other delicate biological mechanisms, such as RNA splicing and the expression of long non-coding RNAs. In this dissertation, we carried out five studies based on high-throughput sequencing data, in an attempt to understand how RNA splicing and differential expression of long non-coding RNAs is associated biological functions. In the first two studies, we identified and characterized 197 stimulant induced and 477 developmentally regulated alternative splicing events from RNA sequencing data. In the third study, we introduced a method for identifying novel alternative splicing events that were never documented. In the fourth study, we introduced a method for identifying known and novel RNA splicing junctions from protein mass spectrometry data. In the fifth study, we introduced a method for identifying long non-coding RNAs from poly-A selected RNA sequencing data. Taking advantage of these methods, we turned RNA sequencing and protein mass spectrometry data into an information gold mine of splicing and long non-coding RNA activities. / 2019-05-06

京都大学 / 0048 / 新制・論文博士 / 博士(医科学) / 乙第13169号 / 論医科博第4号 / 新制||医科||6(附属図書館) / 京都大学大学院医学研究科医科学専攻 / (主査)教授 萩原 正敏, 教授 近藤 玄, 教授 高田 穣 / 学位規則第4条第2項該当 / Doctor of Medical Science / Kyoto University / DFAM

Recent advances in biological research point out that many ribonucleic acids (RNAs) are transcribed from the genome to perform a variety of cellular functions, rather than merely acting as information carriers for protein synthesis. These RNAs are usually referred to as the non-coding RNAs (ncRNAs). The versatile regulation mechanisms and functionalities of the ncRNAs contribute to the amazing complexity of the biological system. The ncRNAs perform their biological functions by folding into specific structures. In this case, the comparative study of the ncRNA structures is key to the inference of their molecular and cellular functions. We are especially interested in two computational problems for the comparative analysis of ncRNA structures: the alignment of ncRNA structures and their classification. Specifically, we aim to develop algorithms to align and cluster RNA structural motifs (recurrent RNA 3D fragments), as well as RNA secondary structures. Thorough understanding of RNA structural motifs will help us to disassemble the huge RNA 3D structures into functional modules, which can significantly facilitate the analysis of the detailed molecular functions. On the other hand, efficient alignment and clustering of the RNA secondary structures will provide insights for the understanding of the ncRNA expression and interaction in a genomic scale. In this dissertation, we will present a suite of computational algorithms and software packages to solve the RNA structural motif alignment and clustering problem, as well as the RNA iii secondary structure alignment and clustering problem. The summary of the contributions of this dissertation is as follows. (1) We developed RNAMotifScan for comparing and searching RNA structural motifs. Recent studies have shown that RNA structural motifs play an essential role in RNA folding and interaction with other molecules. Computational identification and analysis of RNA structural motifs remain to be challenging tasks. Existing motif identification methods based on 3D structure may not properly compare motifs with high structural variations. We present a novel RNA structural alignment method for RNA structural motif identi- fication, RNAMotifScan, which takes into consideration the isosteric (both canonical and non-canonical) base-pairs and multi-pairings in RNA structural motifs. The utility and accuracy of RNAMotifScan are demonstrated by searching for Kink-turn, C-loop, Sarcin-ricin, Reverse Kink-turn and E-loop motifs against a 23s rRNA (PDBid: 1S72), which is well characterized for the occurrences of these motifs. (2) We improved upon RNAMotifScan by incorporating base-stacking information and devising a new branch-and-bound algorithm called RNAMotifScanX. Model-based search of RNA structural motif has been focused on finding instances with similar 3D geometry and base-pairing patterns. Although these methods have successfully identified many of the true motif instances, each of them has its own limitations and their accuracy and sensitivity can be further improved. We introduce a novel approach to model the RNA structural motifs, which incorporates both base-pairing and base-stacking information. We also develop a new algorithm to search for known motif instances with the consideration of both base-pairing and base-stacking information. Benchmarking of RNAMotifScanX on searching known RNA structural motifs including kink-turn, C-loop, sarcin-ricin, reverse kink-turn, and E-loop iv clearly show improved performances compared to its predecessor RNAMotifScan and other state-of-the-art RNA structural motif search tools. (3) We develop an RNA structural motif clustering and de novo identification pipeline called RNAMSC. RNA structural motifs are the building blocks of the complex RNA architecture. Identification of non-coding RNA structural motifs is a critical step towards understanding of their structures and functionalities. We present a clustering approach for de novo RNA structural motif identification. We applied our approach on a data set containing 5S, 16S and 23S rRNAs and rediscovered many known motifs including GNRA tetraloop, kink-turn, C-loop, sarcin-ricin, reverse kink-turn, hook-turn, E-loop and tandem-sheared motifs, with higher accuracy than the currently state-of-the-art clustering method. More importantly, several novel structural motif families have been revealed by our novel clustering analysis. (4) We propose an improved RNA structural clustering pipeline that takes into account the length-dependent distribution of the structural similarity measure. We also devise a more efficient and robust CLique finding CLustering algorithm (CLCL), to replace the traditional hierarchical clustering approach. Benchmark of the proposed pipeline on Rfam data clearly demonstrates over 10% performance gain, when compared to a traditional hierarchical clustering pipeline. We applied this new computational pipeline to cluster the posttranscriptional control elements in fly 3’-UTR. The ncRNA elements in the 3’ untranslated regions (3’-UTRs) are known to participate in the genes’ post-transcriptional regulation, such as their stability, translation efficiency, and subcellular localization. Inferring co-expression patterns of the genes by clustering their 3’-UTR ncRNA elements will provide invaluable knowledge for further studies of their functionalities and interactions under specific physiological processes. v (5) We develop an ultra-efficient RNA secondary structure alignment algorithm ERA by using a sparse dynamic programming technique. Current advances of the next-generation sequencing technology have revealed a large number of un-annotated RNA transcripts. Comparative study of the RNA structurome is an important approach to assess the biological functionalities of these RNA transcripts. Due to the large sizes and abundance of the RNA transcripts, an efficient and accurate RNA structure-structure alignment algorithm is in urgent need to facilitate the comparative study. By using the sparse dynamic programming technique, we devised a new alignment algorithm that is as efficient as the tree-based alignment algorithms, and as accurate as the general edit-distance alignment algorithms. We implemented the new algorithm into a program called ERA (Efficient RNA Alignment). Benchmark results indicate that ERA can significantly speedup RNA structure-structure alignments compared to other state-of-the-art RNA alignment tools, while maintaining high alignment accuracy. These novel algorithms have led to the discovery of many novel RNA structural motif instances, which have significantly deepened our understanding to the RNA molecular functions. The genome-wide clustering of ncRNA elements in fly 3’-UTR has predicted a cluster of genes that are responsible for the spermatogenesis process. More importantly, these genes are very likely to be co-regulated by their common 3’-UTR elements. We anticipate that these algorithms and the corresponding software tools will significantly promote the comparative ncRNA research in the future

Long non-coding RNAs (lncRNAs) have been implicated as important regulators of inflammation through various mechanisms in both the innate and adaptive immune systems of mice and humans. The majority of SNPs identified by GWAS to be associated with autoimmune disorders lie within non-coding areas of the genome, including genes for lncRNAs. To identify lncRNAs with relevancy to inflammation and autoimmunity, a discovery pipeline was used to find lncRNAs differentially expressed in TLR4 activated murine macrophages, conserved between mice and humans, and harboring GWAS identified SNPs associated with autoimmune disorders. Two of the six candidate lncRNAs identified, Lnc15 and Lnc13, are decreased in activated macrophages and are associated with both celiac and Crohn’s disease. To further explore the regulation and influence of these two lncRNAs during inflammation and its resolution, a variety of in vitro and in vivo techniques were utilized, including novel mouse knockout models. An investigation of Lnc15 was conducted in cells of both the innate and adaptive immune system, where the dominant isoform of Lnc15 was identified to be a ~1.4 kb transcript localized to the cytoplasm in both murine macrophages and T cells. Analysis of Lnc15 regulation was conducted in activated murine macrophages, focused on TLR signaling. Through stimulating macrophages with specific TLR ligands, Lnc15 was found to be decreased by TLR2, TLR3, and TLR4 signaling, likely dependent upon both MYD88 and TRIF. While not dependent upon NF-κB, protein synthesis is required for TLR induced decreases in Lnc15 levels. Conversely, activated neutrophils significantly increase Lnc15 levels, although the mechanism of regulation is not yet known. Mice lacking Lnc15 globally were found to be more susceptible to DSS induced colitis, which is likely dependent upon a defect in the innate immune system. In the adaptive immune system, Lnc15 was found to be specifically upregulated in Tregs compared to other T cell subsets. Lnc15 deficient Tregs had a reduced suppressive capacity in vitro, but not in vivo in a T cell induced model of colitis. These findings suggest Lnc15 plays a role in Treg suppressive capacity under certain conditions, but the exact mechanism influenced remains to be identified. Additionally, overexpression of Lnc15 in a murine T cell line resulted in a decrease in Rorc expression. A Lnc15 RNA pulldown experiment identified USF2, a transcription factor known to regulate Rorc expression, and UBR5, a ubiquitin-protein ligase known to influence RORyt stability, as protein interactors of Lnc15. These data indicate that Lnc15 can influence aspects of RORyt biology, which implicates Lnc15 as a regulator of either the plasticity between Tregs and Th17 cells, or Treg ability to suppress inflammatory Th17 cells. An investigation into Lnc13 regulation by disease relevant cytokines was conducted with a series of macrophage stimulation experiments. Lnc13 was found to be positively regulated by cytokines with an anti-inflammatory capacity, including IL-6, IL-4 and IL-10. When Lnc13 deficient macrophages were polarized, a higher expression of Il6 was detected in both M1 and M2 macrophages, suggesting a regulatory connection between Lnc13 and IL-6 across macrophage activation states. Previously identified Lnc13 target genes displayed a quicker transcriptional response to LPS stimulation in Lnc13 deficient macrophages. Additionally, when the Lnc13 mouse was crossed with the DQ8 transgenic mouse model and challenged to gluten, the ileum tissue of Lnc13 deficient mice expressed higher amounts of Il12 and Ifng, cytokines directly relevant to celiac disease. These findings provide support for Lnc13 as a novel regulator of macrophage response and cytokine expression in response to disease relevant stimuli.

Work over the past two decades has revealed that non-coding RNAs (ncRNAs) are prevalent in all kingdoms of life. Using RNA-seq we discovered hundreds of ncRNAs in the antibiotic-producing genus of bacteria, Streptomyces. These included trans-encoded small RNAS (sRNAs), cis-antisense RNAs, and a new type of antisense RNA we termed cutoRNAs (convergent untranslated overlapping RNAs) that arise when transcription termination does not occur in the intergenic region between two convergently arranged genes. Many of these ncRNAs feature prominently in the specialized metabolite biosynthetic clusters (e.g. antibiotics, anticancer agents, immunosuppressants). Hence, it is likely that understanding the functions of these RNAs will be important for new molecule discovery. We found that one highly expressed antisense RNA (ScbN) was expressed opposite the -butyrolactone synthase scbA in the model streptomycete Streptomyces coelicolor. However, ScbN had no detectible impact on the expression of scbA. Instead, the transcription terminator of scbN, which also forms a hairpin within the coding sequence of scbA, was found to reduce expression of scbA more than 10-fold. This led us to bioinformatically search for similar coding-sequence hairpins throughout all bacteria, leading to the discovery of many stable RNA structures with conserved locations throughout very divergent bacteria (e.g. Streptomyces, Escherichia coli, Bacillus subtilis). / Thesis / Doctor of Philosophy (PhD) / The flow of genetic information, from DNA to RNA to proteins, often portrays RNA as a mere intermediary molecule. An alternative, and perhaps more accurate, way to view RNA is that it is central to all cellular processes. Many RNAs are not translated into proteins and instead act as regulatory molecules, impacting the expression of other genes. In this work we found many examples of these regulatory RNAs in a group of bacteria known to produce many of the world’s antibiotics. Understanding the roles these regulatory RNAs play in impacting gene expression will be important for the discovery of new molecules, such as antibiotics. In addition to distinct regulatory RNAs mentioned above, we found that RNA structures within the coding sequences of mRNAs that are translated into proteins have dramatic regulatory consequences. We describe the characterization of one such RNA structure in a gene involved in bacterial communication, and develop a bioinformatic tool to hunt for other such structures conserved throughout bacteria.

Prader-Willi Syndrome (PWS) is a genetic disease that results in abnormal hormone levels, developmental delay, intellectual disability, hypogonadism, and excessive appetite. The disease is caused by a de novo genetic deletion in chromosome 15. While many of the deleted genes have been identified, there is little known about their molecular function. There is evidence that a cluster of non-coding RNA genes in the deleted region known as the SNORD116 genes may be the most critical genes deleted in Prader-Willi Syndrome. It is unknown what the SNORD116 genes do at the molecular level, but recent evidence suggests they regulate the expression of other genes involved in the neuroendocrine system. Specifically, the SNORD116 gene is implicated in regulation of NHLH2, a transcription factor gene which plays a key role in development, hormonal regulation, and body weight. In this study we identify phylogenetically conserved regions of SNORD116 and predict interactions with its potential downstream RNA targets. We show that mouse Snord116 post-transcriptionally increases Nhlh2 RNA levels dependent on its 3'UTR and protects it from degradation within 45 minutes of its transcription. Additionally, a single nucleotide variant within Nhlh2 at the predicted Snord116 interaction site may disrupt Snord116's protective effect. This is the first observation of a molecular mechanism for Snord116, identifying its role in RNA stability, and leads us closer to understanding Prader-Willi Syndrome and finding a possible treatment. However, Snord116 in vitro knockdown or paternally inherited in vivo deletion fail to detect differential expression of Nhlh2, likely due to missing the key timepoint of Snord116 regulatory effects on Nhlh2 RNA soon after its transcriptional stimulation, and dependent on leptin signals. Furthermore, the hypothalamic mRNA expression profile of PWS mouse models fed a nutraceutical dietary supplement of conjugated linoleic acid reveals minimal overall changes, while the effect of diet may be stronger than genotype and potentially changes gene expression of metabolic molecular pathways. / Doctor of Philosophy / Prader-Willi Syndrome is a genetic disease that results in abnormal hormone levels, slow development, intellectual disability, gonad deficiency, and excessive appetite. The disease is caused by a genetic deletion in chromosome 15 that is almost always a spontaneous mutation not inherited from the parents. While many of the deleted genes have been identified, there is little known about what their molecular function is. There is evidence that a cluster of genes in the deleted region known as the SNORD116 genes may be the most critical genes deleted in Prader-Willi Syndrome. It is unknown what the SNORD116 genes do at the molecular level, but recent evidence suggests that it regulates other genes involved in the hormone system. Specifically, the SNORD116 gene is implicated to regulate the levels of NHLH2, a gene which plays a key role in development, hormonal regulation, and body weight. In this study we identify key regions of SNORD116 and predict interactions with its potential downstream targets. We show that SNORD116 increases NHLH2 levels and slows its degradation at the RNA transcript level. This is the first observation of a molecular mechanism for SNORD116 and leads us closer to understanding Prader-Willi Syndrome and finding a possible treatment. However, other mouse models of Snord116 deletion fail to find differences in Nhlh2. This is likely due to missing a brief key timepoint and hormonal signal when Nhlh2 is most subject to Snord116's effects. Furthermore, PWS mouse models fed a supplement intended for weight loss leads to mild overall gene expression changes in the hypothalamus, a brain region that regulates many hormonal signals including appetite and energy balance. The effect of diet may be stronger than genotype in this brain region, with diet potentially changing the activity of metabolic molecular pathways.

The objective of this research is to develop a novel computational prediction algorithm for non-coding RNA (ncRNA) genes using features computable for any genomic sequence without the need for comparative analysis. Existing comparative-based methods require the knowledge of closely related organisms in order to search for sequence and structural similarities. This approach imposes constraints on the type of ncRNAs, the organism, and the regions where the ncRNAs can be found. We have developed a novel approach for ncRNA gene prediction without the limitations of current comparative-based methods. Our work has established a ncRNA database required for subsequent feature and genomic analysis. Furthermore, we have identified significant features from folding-, structural-, and ensemble-based statistics for use in ncRNA prediction. We have also examined higher-order gene structures, namely operons, to discover potential insights into how ncRNAs are transcribed. Being able to automatically identify ncRNAs on a genome-wide scale is immensely powerful for incorporating it into a pipeline for large-scale genome annotation. This work will contribute to a more comprehensive annotation of ncRNA genes in microbial genomes to meet the demands of functional and regulatory genomic studies.

Since their discovery in 1953, adenoviruses have significantly contributed to the understanding of virus-host cell interactions, including mechanistic details of cellular processes such as cell cycle control and alternative RNA splicing. Among the first characterized adenoviral genes were the virus-associated RNAs (VA RNAI/II), which are produced in massive amount during a lytic infection. The VA RNAs perform multiple functions and are required for a successful adenovirus life cycle. More recently, it was shown that the VA RNAs are processed into small viral miRNAs, so-called mivaRNAs, which interfere with the function of the cellular RNAi/miRNA machinery. In papers I and II, we focused on a structural and functional characterization of the mivaRNAs using two approaches. Firstly, we created a model system where the predicted miRNA-like function of mivaRNAI could be investigated, without interfering with other VA RNA functions. This was accomplished by construction of recombinant adenoviruses, in which the seed sequence of mivaRNAI was altered. The results showed that in cell culture experiments the mivaRNAI seed sequence mutants grew as the wild type virus, suggesting that the mivaRNAs are not required during the lytic phase of an adenovirus infection. Secondly, we showed that the VA RNAs from different human adenoviruses (Ad4, Ad5, Ad11 and Ad37) undergo the same type of Dicer-dependent processing into mivaRNAs, which subsequently are loaded onto the RNA induced silencing complex (RISC), albeit with different efficiencies. In paper III, we demonstrated that the promoter proximal region of the adenovirus major late promoter (MLP) produces a novel non-canonical class of small RNAs, which we termed the MLP-TSS-sRNAs. Surprisingly the MLP-TSS-sRNA maintains the m7G-cap structure while bound to Ago2 containing RISC. These complexes are functional suppressing expression of target mRNAs with complementary binding site. Most importantly, the MLP-TSS-sRNA limits the efficiency of viral DNA replication probably through a targeting of the E2B mRNAs, which are transcribed in the antisense orientation. In conclusion, the MLP-TSS-sRNA represents the first viral small RNA, which has been shown to have a function as a regulator of an adenovirus infection.

Myc deregulation critically contributes to many cancer etiologies. Recent work suggests that Myc and its direct interactors can confer a distinct epigenetic state. Our goal is to better understand the Myc-conferred epigenetic status of cells. We have previously identified the long non-coding RNA (lncRNA), H19, as a target of Myc regulation and shown it to be important for transformation in lung and breast cells. These results prompted further analysis to identify similarly important Myc-regulated lncRNAs. Myc-regulated lncRNAs associated with the cell cycle and transformation have been identified by microarray analysis. A small number of candidate lncRNAs that were differentially expressed in both the cell cycle and transformation have been validated. Given the increasing importance of lncRNAs and epigenetics to cancer biology, the discovery of Myc-induced, growth associated lncRNAs could provide insight into the mechanisms behind Myc-related epigenetic signatures in both normal and disease states.