Computational modelling of the 3D structure of RNA - small molecule ligand complexes

背景：核糖核酸（RNA）是在所有生命形式中发现的重要分子。它们的作用范围从遗传信息的载体（即翻译成蛋白质的信使RNA）到基因表达的调节剂（例如微RNA）或生化反应的催化剂（例如核酶）。到目前为止，我们估计在细胞中转录的高达97%的RNA不用于编码蛋白质。通常，这些非编码RNA（ncRNA）使用复杂的结构来实现其非编码功能。 RNA结构是分层组织的，可以在多个层次上描述。首先，二级结构是腺嘌呤（A）和尿嘧啶（U）、鸟嘌呤（G）和胞嘧啶（C）或鸟嘌呤和尿嘧啶碱基之间的所有典型碱基配对相互作用的集合。尽管如此，在二级结构的未配对区域中也发现了其他类型的碱基对，在那里它们形成用于稳定这些局部基序的3D结构的复杂网络。我们称之为2.5D结构，即由规范碱基对堆叠而成并由局部3D基序连接的茎的集合。这种2.5D结构形成了分子完整3D结构的支架，我们将其表示为三级结构。 折叠的结构和动力学是用来控制分子功能的重要性质。特别地，另一种分子（例如蛋白质、RNA、核苷酸）结合靶RNA以改变其构象并激活其分子功能。最近的研究发现，小分子在许多细胞途径中作为RNA功能的重要非共价调节剂。这些发现有助于更好地理解调节生物系统的分子机制，但也使RNA分子成为一大类有前途的新型药物靶标，其应用包括新型抗生素，抗病毒药和CRISPR激活剂。 贡献：该提案旨在开发预测RNA 2.5D和3D结构的算法，并利用这些信息来预测小分子结合RNA。这些方法将使我们能够分析小分子结合对RNA稳定性的影响，并检测由这种结合引发的推定的结构重组。 RNA 3D基序的自动识别：我们将开发图形算法来提取经常性的局部3D基序，并利用这些信息来预测序列数据的发生。 RNA 2.5D结构预测：使用动态规划和整数规划技术，我们将联合收割机经典的二级结构预测与我们的RNA 3D基序预测相结合，以预测2.5D和3D结构。 小分子结合预测：我们将开发图形神经网络来学习小分子与RNA 3D基序的结合偏好。并利用这项技术注释细菌基因组中的药物靶点。 RNA适体的计算设计：我们将建立算法，用于设计对靶小分子具有高结合亲和力的RNA。

BACKGROUND: Ribonucleic Acids (RNAs) are essential molecules found in all forms of life. Their role ranges from carrier of the genetic information (i.e. messenger RNA that are translated into proteins) to regulator of gene expression (e.g. micro RNA) or catalyzer of biochemical reactions (e.g. ribozymes). To date, we estimate that up to 97% of the RNA transcribed in cells is not used to encode proteins. Often, these non-coding RNAs (ncRNAs) use sophisticated structures to fulfil their non-coding functions. RNA structures are hierarchically organized and can be described at multiple levels. First, the secondary structure is the set of all canonical base-pairing interactions between Adenine (A) and Uracil (U), Guanine (G) and Cytosine (C), or Guanine and Uracil bases. Though, other types of base pairs are also found in unpaired regions of the secondary structure, where they form complex networks used to stabilize the 3D structures of these local motifs. We call a 2.5D structure the collection of stems built from stacks of canonical base pairs and connected by local 3D motifs. This 2.5D structure forms the scaffold of the complete 3D structure of the molecule that we denote as the tertiary structure. The structure but also the dynamics of the folding are important properties used to control molecular functions. In particular, another molecule (e.g. protein, RNA, nucleotide) binds the target RNA to change its conformation and activate its molecular function. Recent studies identified small molecules as important non-covalent regulators of RNA function in many cellular pathways. These discoveries contribute to a better understanding of molecular mechanisms regulating biological systems, but also pose RNA molecules as a large class of promising novel drug targets, with applications including novel antibiotics, antivirals, and CRISPR activators. CONTRIBUTION: This proposal aims to develop algorithms for predicting RNA 2.5D and 3D structures and leverage this information to predict small molecules binding RNAs. These methods will allow us to analyze the impact of the binding of a small molecule on the stability of RNAs and detect putative structural reorganizations triggered by this binding. Automated identification of RNA 3D motifs: We will develop graph algorithms to extract recurrent local 3D motifs and leverage this information to predict occurrence from sequence data. RNA 2.5D structure prediction: Using dynamic programming and integer programming techniques, we will combine classical secondary structure predictions with our RNA 3D motif predictions to predict 2.5D and 3D structures. Small-molecule binding prediction: We will develop graph neural networks to learn small molecule binding preferences to RNA 3D motifs. And use this technology to annotate drug targets in bacterial genomes. Computational design of RNA aptamers: We will build algorithms for designing RNAs with high binding affinities to target small molecules.