Advanced algorithms to infer and analyze 3D genome structures

用于推断和分析 3D 基因组结构的先进算法

• 批准号: 10027542
• 负责人: Zheng Wang
• 金额: $ 35.47万
• 依托单位: UNIVERSITY OF MIAMI CORAL GABLES
• 依托单位国家: 美国
• 财政年份: 2020
• 资助国家: 美国
• 起止时间: 2020-09-01 至 2025-08-31
• 项目状态: 未结题
• 来源: https://reporter.nih.gov/project-details/10027542
• 关键词: 3-Dimensional; Address; Algorithms; Bioinformatics; Biological; Cells; Communities; Complex; DNA; Data; Databases; Disease; Enhancers; Epigenetic Process; Gene Expression Regulation; Genetic; Genome; Goals; Individual; Malignant Neoplasms; Neurons; Other Genetics; Play; Population; Positioning Attribute; Process; Productivity; Proteins; Regulatory Element; Research; Resolution; Rewards; Role; Scientist; Structure; Techniques; Untranslated RNA; Work; base; chromatin remodeling; cohesin; deep learning algorithm; genome database; genome-wide; improved; knowledge base; member; promoter; protein complex; recruit; scaffold; three dimensional structure; whole genome

Project Summary For the past decade, the population-cell Hi-C technique has significantly improved our ability to discover genome-wide DNA proximities. However, because population Hi-C is based on a pool of cells, it will not help us reveal each single cell's 3D genome structure or understand cell-to-cell variability in terms of 3D genome structure and gene regulation. It is also difficult to achieve a high resolution, such as 1 Kbp, with population Hi- C; therefore, when finding and analyzing the spatial interactions for the promoter or enhancer regions typically associated with biologically-important regulatory elements, population Hi-C data's resolution is too low to be useful. Moreover, while we know that the CTCF-cohesin complex plays a key role in the formation of genome 3D structures, the question is whether long non-coding RNAs (lncRNAs) are involved in the process since lncRNAs have been found to recruit proteins needed for chromatin remodeling, and our preliminary research has found that lncRNA LINC00346 directly interacts with CTCF. Finally, while members of the bioinformatics community, including the PI, have developed many algorithms to reconstruct 3D genome structures based on population Hi-C data, important questions still must be answered regarding how 3D genome structures are involved in gene regulation and whether there are relationships between 3D genome structures and genetic and epigenetic features. The PI proposes to conduct leading research to overcome these challenges and address these questions. During the next five years, the PI will develop algorithms to reconstruct the 3D whole- genome structures for single cells and analyze cell-to-cell variabilities in terms of 3D genome structure and gene regulation. The PI will develop a deep learning algorithm to enhance the resolution of population Hi-C data to that of Capture Hi-C data (1 Kbp) so that we can make good use of the large amount of Hi-C data accumulated in the past decade. An online database will be built to allow the community to access both population and single-cell 3D genome structures in an integrated way. The PI will work with a cancer biologist to discover any lncRNAs that function as a scaffold to fine-tune the CTCF-cohesin protein complex, as well as two neuron scientists to develop a more complete understanding of gene regulation while considering 3D genome and other genetic and epigenetic features. Given the PI's track record and productivity, having three computational goals and two collaborative goals is not only feasible but computationally and biologically rewarding. In five years, once the proposed studies are accomplished, the PI should have established a uniquely independent place in the field of 3D genome, maintaining leading positions in inferring single-cell 3D genome structures, enhancing Hi-C data resolution, and building 3D genome databases, while establishing similar positions in reconstructing high-resolution 3D genome structures, finding lncRNAs' roles in the formation of genome structures, and understanding how 3D genome structures are involved in gene regulation.

项目概要 在过去的十年中，群体细胞 Hi-C 技术显着提高了我们发现 全基因组 DNA 邻近性。然而，由于群体 Hi-C 是基于细胞池的，因此这无济于事 我们揭示每个单细胞的 3D 基因组结构或了解 3D 基因组方面的细胞间变异 结构和基因调控。实现高分辨率（例如 1 Kbp）也很困难，因为群体 Hi- C;因此，在寻找和分析启动子或增强子区域的空间相互作用时，通常 与生物学上重要的调控元件相关，群体 Hi-C 数据的分辨率太低，无法 有用。此外，虽然我们知道 CTCF-cohesin 复合物在基因组的形成中起着关键作用 3D 结构，问题是长非编码 RNA (lncRNA) 是否参与该过程，因为 研究发现lncRNA可以招募染色质重塑所需的蛋白质，我们的初步研究 发现lncRNA LINC00346直接与CTCF相互作用。最后，虽然生物信息学成员 包括 PI 在内的社区已经开发了许多算法来重建 3D 基因组结构 人口 Hi-C 数据，仍然必须回答关于 3D 基因组结构如何的重要问题 参与基因调控以及3D基因组结构与遗传之间是否存在关系 和表观遗传特征。 PI 建议开展领先研究来克服这些挑战 解决这些问题。在接下来的五年中，PI 将开发算法来重建 3D 整体 单细胞的基因组结构，并根据 3D 基因组结构分析细胞间的变异 基因调控。 PI将开发深度学习算法来提高群体Hi-C的分辨率 数据到Capture Hi-C数据（1 Kbp），以便我们可以充分利用大量的Hi-C数据 过去十年的积累。将建立一个在线数据库，以允许社区访问 以集成的方式研究群体和单细胞 3D 基因组结构。 PI 将与癌症生物学家合作 发现任何作为支架来微调 CTCF-粘连蛋白复合物的 lncRNA，以及 两位神经元科学家在考虑 3D 的同时对基因调控有更全面的了解 基因组和其他遗传和表观遗传特征。鉴于 PI 的业绩记录和生产力，拥有三个 计算目标和两个协作目标不仅是可行的，而且在计算和生物学上都是可行的 有益的。五年后，一旦完成拟议的研究，PI 应建立一个 在3D基因组领域拥有独特的独立地位，在单细胞3D推断领域保持领先地位 基因组结构，增强 Hi-C 数据分辨率，构建 3D 基因组数据库，同时建立 重建高分辨率 3D 基因组结构中的相似位置，发现 lncRNA 在 基因组结构的形成，并了解 3D 基因组结构如何参与基因调控。