Computational Approaches for RNA StructureFunction Determination

-计算检测基因组中丰富的远程核苷酸共变异-开发了一种新的计算方法来检测基因组尺度上的远程共变异，产生高和中等置信度的相互作用网络。为了测试该算法，将其应用于一组对齐的果蝇基因组，发现mrna和非编码rna（包括内切sirna）的外显子之间存在普遍的远程共变异。大多数共变异似乎是同步突变的结果，而不是反向互补。发现共变的基因丰富了与区域化以及神经和发育过程相关的功能。在较低置信度的相互作用网络中，存在对具有细胞成分注释（如质膜或细胞骨架）的基因的偏见。这可能是搭便车现象的一个普遍迹象，在搭便车现象中，转录本与其他rna结合，这些rna是沿细胞骨架运输的核糖核蛋白复合物的一部分。这些相互作用可能以统计大小依赖的方式发生。这些结果表明，RNA-RNA远程相互作用是一种广泛的现象，对各种细胞过程都很重要。-一种新的翻译增强子的发现和表征-细胞和病毒mrna的3' utr含有通过使用未知机制增强翻译而在基因表达中起作用的元件。为了确定这些元素的功能，我们之前使用了萝卜皱病毒（TCV）。TCV以帽独立的方式翻译，包含一个3‘区域，与5’ UTR一起协同增强翻译。我们最近在豌豆代花叶病毒中发现了另一种这种类型的元件，然而，这种翻译增强元件以某种不同的方式起作用。我们使用了一套计算工具，包括我们实验室开发的用于3D RNA建模的RNA2D3D，以及实验方法来证明该元素是t形的，这让人想起TCV与tRNA的相似性。我们发现它与核糖体结合并参与长距离的接吻环RNA-RNA相互作用。它的功能提示了RNA 5‘/3’环化的一种手段，因此是一种增强翻译的机制。这些新元素的存在似乎暗示了在真核生物中可能发现的替代翻译机制。-了解登革热病毒的结构和功能-最近据报道，世界上有4亿多登革热病例，比以前报道的多4倍，其中10%导致严重形式的疾病。利用我们的RNA折叠大规模并行遗传算法，我们发现登革热病毒RNA的3'非翻译区核心区域可以形成2个出现频率不等的哑铃结构。实验表明，由这些哑铃形成的结构基序对病毒复制很重要。结果表明，两个哑铃在翻译过程中具有协同效应。因此，登革病毒核心区的顺式作用元件是复制和最佳翻译所必需的。最近，我们使用酰基化技术（SHAPE）、铅诱导水解和定点诱变对登革热微小基因组进行了详细的结构分析。结果表明，5‘和3’末端区域之间的蛋白独立相互作用证实了许多预测的相互作用，并揭示了在3‘ UTR中发现的假结相互作用和序列依赖于5’环化区域。-离子和侧翼碱基在HIV-1二聚化中的作用-实验表明，HIV-1接吻环形成的特征取决于病毒的亚型。研究表明，b亚型单体在高盐浓度或镁离子存在下二聚化，而a亚型单体只与镁离子结合在G273或G274的磷酸基上二聚化，而与盐浓度无关。我们通过计算机模拟发现，在低浓度下，两种类型的单体发夹环都明显变形，发夹环中与二聚化有关的碱基向内转动。在高盐浓度下，b亚型单体保持有利于二聚化的形状，而a亚型单体仍表现出明显的变形。b亚型的侧翼碱基有助于稳定构象，而a亚型的侧翼碱基G273引起变形。然而，当镁离子存在并与G273或G274磷酸基团结合时，碱G273保持稳定环二聚化的构象。这些结果对于理解HIV-1病毒生命周期的机制非常重要。microrna是非编码rna，由来自前体分子的约22个核苷酸组成。这些前体通常折叠成茎环发夹结构。在扫描基因组时，很难将假阳性的pre- mirna与真实的区分开来。本研究提出了一种鉴定和分类pre- mirna的新方法。采用组合特征挖掘的方法来发现一组好的特征。然后使用这些特征集来训练支持向量机以获得分类模型。然后应用增强算法进一步提高精度。结果表明，与以往的方法相比，有显著的改进。-发现两个RNA序列中的共同折叠模式-开发了一种高效的动态规划算法，该算法使用有序标记树来发现给定2个RNA序列中最大的共同RNA亚结构。该算法还可用于发现RNA二级结构的重复区域。通过李群统一的DNA和RNA双螺旋多尺度建模-为了进一步理解核酸螺旋固有的运动，综述了双链核酸螺旋结构（DNA和RNA）的连续和离散力学建模的几种方法，并采用了一种新的李群分析方法来调和模型。对刚体运动进行了分析。在最粗的水平上，采用具有各向异性弯曲刚度的蠕虫状链。然后证明了双棒模型在足够长的长度内收敛于此。然后，我们展示了原子力显微镜获得的全原子分辨率分子动力学和实验结果与模型的关系。

-Computational Detection of Abundant Long-range Nucleotide Co-variation in Genomes- A novel computational methodology was developed to detect long-range co-variation on a genomic scale producing high and medium confidence interactinos networks. To test the algorithm, it was applied to a set of aligned drosophila genomes where it was found that there are prevalent long-range co-variations between exons of mRNAs and non-coding RNAs, including endo-siRNAs. Most of the co-variations appear to be the result of synchronized mutations, not reverse complementarity. The genes found to co-vary are enriched for functions related to regionalization as well as neural and developmental processes. In a lower confidence interaction network there is a bias towards genes that have cellular component annotations such as plasma membrane or cytoskeleton. This may be an indicator of a widespread phenomenon of hitchhiking, in which transcripts bind to other RNAs that are part of ribonucleoprotein complexes which are transported along the cytoskeleton. These interactions may occur in a statistical size dependent manner. These results indicate that RNA-RNA long-range interactions are a widespread phenomenon that is important to a variety of cellular processes. - Discovery and Characterization of a Novel Translational Enhancers - 3' UTRs of cellular and viral mRNAs harbor elements that function in gene expression by enhancing translation using unknown mechanisms. To determine the function of these elements we previously used the Turnip crinkle virus (TCV). TCV is translated in a cap-independent fashion and contains a 3' region that together with the 5' UTR synergistically enhances translation. We recently discovered another element of this type in the Pea enation mosaic virus, however, this translational enhancer element acts in a somewhat different way. We used a set of computational tools, including RNA2D3D for 3D RNA modeling, developed in our lab, and experimental methods to show that this element is T-shaped, which is reminiscent of TCV's similarity to tRNA. We showed that it binds ribosomes and engages in a long range kissing loop RNA-RNA interaction. Its functionality suggests a means for RNA 5'/3' cyclization and thus a mechanism for translational enhancement. It appears that the existence of these novel elements is suggesting alternate translational mechanisms that may be found in eukaryotic organisms. - Understanding the Structure and Function of the Dengue Virus - It has recently been reported that there are over 400 million cases of dengue fever in the world, 4 times more than previously reported with 10% leading to severe forms of the disease. Using our massively parallel genetic algorithm for RNA folding we showed that the core region of the 3' untranslated region of dengue virus RNA can form 2 dumbell structures of unequal frequencies of occurrence. It was experimentally shown that structural motifs formed from these dumbells are important for viral replication. Also it was shown that there is a cooperative synergy with both dumbells for translation. Thus, the cis-acting elements in the core region of dengue virus are required for both replication and optimal translation. Recently, we performed a detailed structural analysis of the dengue minigenome using acylation techniques (SHAPE), lead induced hydrolysis and site directed mutagenesis. Results indicate protein independent interactions between the 5' and 3' terminal regions corroborating many of the predicted interactions and shedding light on the pseudoknot interactions found in the 3' UTR and the sequence dependency on the 5' cyclization region. -The Role of Ions and Flanking Bases in the Dimerization of HIV-1- Experimentally it has been shown that the characterization of HIV-1 kissing loop formation differs depending on the subtype of the virus. It has been shown that subtype-B monomers dimerize at high salt concentrations or in the presence of magnesium ions while subtype-A monomers will only dimerize with a magnesium ion bound to the flanking G273 phosphate group or the phosphate group of G274 regardless of the salt concentration. We found using computer simulations that at low concentrations both types of monomer hairpin loops were significantly deformed and the bases in the hairpin loop that are associated with dimerization were turned inward. At high salt concentrations the subtype-B monomer maintained a shape conducive to dimerization, while subtype-A still showed significant deformations. Also the flanking bases in subtype-B helped to stabilize the conformation while the flanking base G273 in subtype-A caused deformation. However, when magnesium ions were present and bound to the G273 or G274 phosphate groups, base G273 maintained a conformation that stabilized the loop for dimerization. These results are important for understanding the mechanisms that are involved in the HIV-1 virus life cycle. -Classifying Pre-miRNA via Combinatorial Feature Mining and Boosting MicroRNAs are non-coding RNAs consisting of about 22 nucleotides that are derived from precursor molecules. These precursors usually fold into stem-loop hairpin structures. When scanning genomes it is difficult to distinguish false positive pre-miRNAs from the real thing. In this research a new method was developed for identifying and classifying pre-miRNAs. A combinatorial feature mining approach was used to discover a good set of features. These feature sets were then used to train support vector machines to obtain classification models. A boosting algorithm was then applied to further enhance the accuracy. Results indicate significant improvement over previous methods. -Discovering Common Folding Patterns in two RNA Sequences- An efficient dynamic programming algorithm was developed that uses ordered labeled trees for discovering the largest common RNA substructures given 2 RNA sequences. This algorithm can also be used to discover repeated regions of RNA secondary structure. -Multiscale Modeling of Double Helical DNA and RNA Unified through Lie Groups- In order to further understand motions that are inherent in nucleic acid helices, several means of mechanical modeling, continuous and discrete, of double-stranded nucleic acid helical structures (DNA and RNA)were reviewed and a new analysis method using Lie groups was used to reconcile the models. Rigid body motions were analyzed. At the most coarse level worm-like chains were used with anisotropic bending stiffness. It was then shown that bi-rod models converge to this for long enough lengths. We then showed how full atomic resolution molecular dynamics and experimental results obtained from atomic force microscopy relate to the models.