Reconstruction of 3D Genome Architecture from Chromatin Conformation Capture Data

摘要 我们即将进入构象生物学的新时代。基因组构象对于 许多细胞过程，包括基因调控，与某些变化(易位，富含- 是致癌的。虽然最近的分析，特别是Hi-C，已经改变了人们对 染色质结构，即使是更新的技术也有可能显著改进 三维(3D)基因组重建的准确性和分辨率。然而，要完全 实现这一潜力，将需要新的统计方法和算法来操作 作为结果的数据和结构，并整合伴随的生物医学数据。这个项目的目标是 开发这样的方法。当前的发现提供了一个具体的例子，确定了一个 作为一种新的致癌机制的绝缘邻域破坏的实例。而不是一个 在个别情况下，我们将开发方法来检测全基因组候选，并对其进行优先排序， 基于我们之前在3D热点引出方面的工作。特别是，我们将设计出原创的 无需重建的方法，以避免推断建筑中的不确定性。 尽管存在这些不确定性，但重建带来了几个好处。我们将重新部署 设计的分析，与最近的算法进步相结合，以提高重建精度 和决心。多路FISH提供了更丰富的染色质构象成像，使 改进了将Hi-C触点与距离联系起来的传递函数，这是重建的先兆。 以蛋白质为中心的HiChIP在信息性阅读中提供了收益，多读抢救也是如此。组合 这些进展将产生增强的3D基因组重建方法。 自从潜在的Hi-C以来，3D基因组重建的概念本身就受到了质疑 化验是基于大量的细胞群体。多路复用就地Hi-C实现了 数千个单单元数据集，我们将其与新的多轨迹重建相结合 分析细胞间结构异质性的算法。我们还将利用这些数据来开发 基于结构差异的分类器，用于区分细胞类型。 许多Hi-C数据的下游解释都是通过接触的光谱分析得出的 矩阵，尤其是染色质隔间的描绘。频谱摘要有其局限性 包括在高分辨率下的隔室识别、对归一化的敏感性以及 解释变异。我们将评估接触矩阵的谱分析，重点是 近似对3D重建的影响，通过(I)推断距离矩阵，(Ii) 派生重建，以及(Iii)随后的热点检测。

Abstract We are poised to enter a new era of conformational biology. Genome conformation is critical for numerous cellular processes, including gene regulation, with certain alterations (translocations, fu- sions) being oncogenic. While recent assays, notably Hi-C, have already transformed understanding of chromatin architecture, even newer technologies have the potential to dramatically improve accuracy and resolution of three-dimensional (3D) genome reconstructions. However, to fully realize this potential, new statistical methods and algorithms will be required to operate on the resultant data and structures, and to integrate concomitant biomedical data. This project aims at developing such methods. A concrete example is provided by current findings identifying an instance of insulated neighborhood disruption as a novel oncogenic mechanism. Instead of an individual instance, we will develop methods to detect, and prioritize, genome-wide candidates, building on our previous work on 3D hotspot elicitation. In particular, we will devise original reconstruction-free approaches to avert uncertainties in inferring architecture. Despite these uncertainties, reconstructions confer several advantages. We will deploy newly devised assays, in conjunction with recent algorithmic advances, to improve reconstruction accuracy and resolution. Multiplexed FISH provides richer imaging of chromatin conformation, enabling refinement of transfer functions linking Hi-C contacts to distances, a precursor to reconstruction. Protein-centric HiChIP provides gains in informative reads, as does multi-read rescue. Combining these advances will produce enhanced approaches to 3D genome reconstruction. The very notion of ‘a’ 3D genome reconstruction has been questioned since the underlying Hi- C assays are based on large cell populations. Multiplexed in situ Hi-C has enabled generation of thousands of single-cell datasets which we will couple with a new multi-track reconstruction algorithm to dissect inter-cellular structural heterogeneity. We will also use this data to develop classifiers, based on structural differences, for between cell-type discrimination. Much downstream interpretation of Hi-C data has derived from spectral analysis of the contact matrix, especially delineation of chromatin compartments. Spectral summarization has limitations including compartment identification at high resolution, sensitivity to normalization, and extent of explained variation. We will evaluate spectral analysis of contact matrices with emphasis on the impact of approximations on 3D reconstructions, assessed via (i) inferred distance matrices, (ii) derived reconstructions, and (iii) subsequent hotspot detection.