Algorithms for Managing Uncertainty in Chromosome Conformation Capture Data

描述(由申请人提供)：我们建议研究改进的算法，以解决与染色体构象捕获(3C)分析和相关实验相关的几个挑战。这些最近的高通量实验提供了染色质区域之间的成对联系信息，并提供了几种生物基因组的空间组织的一瞥。它们已被用来计算推断染色质结构的三维模型，并假设基因组特征之间的功能空间关系，如共表达基因、调节区及其调控基因、共同断裂点位置等。然而，用于结构建模、将功能与结构联系起来以及用于可视化3C数据的计算工具仍然缺乏。该提案寻求为几个中央3C分析任务开发计算工具。在目标1中，我们建议将采样与优化框架相结合，以建模与3C数据一致的染色质结构群体。这是至关重要的，因为3C在数百万个细胞中提供了不同结构的平均值。在目标2中，我们设计了一些技术来寻找这些集合中的共同和不同的结构特征，比较不同细胞类型的结构(例如，癌症和正常；淋巴母细胞样变和成纤维细胞)，以及更好的技术来识别在统计上具有显著空间位置的基因组基因座。最后，在Aim 3e中，e提议开发一种“空间基因组浏览器”，它集成了一维基因组注释(基因、甲基化、DNA酶可获得性等)。利用3C型空间数据。我们将应用这些技术来量化人类、酵母和小鼠的细胞间和细胞类型的变异量。使用改进的模型群体，我们将识别新的长程调控实例，并解释现有的假设的远端增强子-启动子相互作用。我们还将把结构与eQTL和GWAS鉴定的SNPs联系起来，以解释eQTL的产生机制和SNP的影响。最后，我们将寻找共表达基因和空间邻近性之间的关系。我们提出的技术将导致更好的结构模型，更有效地计算，并更好地理解结构和功能之间的关系。

DESCRIPTION (provided by applicant): We propose to investigate improved algorithms for several challenges associated with chromosome conformation capture (3C) assays and related experiments. These recent high-throughput experiments give pairwise contact information between chromatin regions and have provided glimpses of the spatial organization of the genomes of several organisms. They have been used to computationally infer three-dimensional models of chromatin structure and to hypothesize functional spatial relationships among genomic features such as co-expressed genes, regulatory regions and their regulated genes, common breakpoint locations, and others. However, computational tools for structural modeling, relating function to structure, and for visualizing 3C data are still lacking. This proposal seeks to develop computational tools for several central 3C analysis tasks. In Aim 1, we propose coupling sampling with an optimization framework to model populations of chromatin structures that are consistent with 3C data. This is essential because 3C provides an average over different structures in millions of cells. In Aim 2, we devise techniques to find common and different structural features within these ensembles, comparing structures of different cell types (e.g. cancer vs. normal; lyphoblastoid vs. fibroblast) and better techniques t identify genomic loci that are statistically significantly spatially co-located. Finally, in Aim 3 e propose to develop a "spatial genome browser" that integrates both 1-d genomic annotations (genes, methylation, DNAase accessibility, etc.) with 3C spatial data. We will apply these techniques to quantifying the amount of cell-to-cell and cell-type variation in human, yeast, and mouse. Using improved populations of models, we will identify new instances of long-range regulation and explain existing postulated distal enhancer-promoter interactions. We will also correlate structure with eQTLs and GWAS-identified SNPs to explain the mechanism causing the eQTL and the effect of the SNP. Finally, we will search for relationships between co-expressed genes and spatial proximity. The techniques we propose will result in better structural models computed more efficiently and a better understanding of the relationships between structure and function.