Identification of methylation haplotype blocks aids in deconvolution of heterogeneous tissue samples and tissue-of-origin mapping from plasma DNA [Tumor RRBS]

Adjacent CpG sites in mammalian genomes tend to be co-methylated due to the processivity of enzymes responsible for adding or removing the methyl group. Yet discordant methylation patterns have also be observed, and found to be related to stochastic or uncoordinated molecular processes. Here we focused on a systematic search and investigation of regions in the human genome that exhibit highly coordinated methylation. By examining the co-methylation patterns of multiple adjacent CpG sites, termed methylation haplotypes, in single bisulfite sequencing reads, we applied a greedy-searching strategy to defined blocks of tightly coupled CpG sites, called Methylation Haplotype Blocks (MHBs), based on 53 sets of whole genome bisulfite sequencing (WGBS) data, including 43 published sets from human adult tissues, ESC and in vitro differentiated cell lines, as well as 10 sets from human adult tissues generated in this study. The MHBs were then further validated with 101 sets of RRBS ENCODE data, and 637 sets of Illumina 450k methylation array data from TCGA tumor and normal samples. Globally, MHBs are enriched in but only partially overlap with several well-known genomic features, including CpG islands, promoters, enhancers and VMRs. To perform quantitative analysis of the MHBs, we defined a metric called Methylation Haplotype Load (MHL), which is covered both average methylation level and methylation complexity and therefore more informative than average methylation level or Shannon entropy. Using a feature selection strategy, we identified a set of tissue-specific MHBs that cluster by developmental germ-layers. Interestingly, examination of these MHBs revealed two distinct mechanisms for fate commitment during development: epigenetic silencing of pluripotent genes, such as NANOG, for mesoderm induction; and epigenetic induction (or de-suppression) of lineage-specific factors for ectoderm commitment. Examination of co-methylation in primary tumor tissues

哺乳动物基因组中相邻的CpG位点（CpG site）往往会发生共甲基化，这是由于负责添加或去除甲基基团的酶所具备的持续合成活性所致。然而，研究人员也已观测到不一致的甲基化模式，且发现其与随机或不协调的分子过程相关。本研究聚焦于人类基因组中呈现高度协同甲基化的区域的系统性搜寻与分析。通过对单分子亚硫酸氢盐测序读段中多个相邻CpG位点的共甲基化模式——即甲基化单倍型（methylation haplotype）——进行分析，我们基于53组全基因组亚硫酸氢盐测序（whole genome bisulfite sequencing, WGBS）数据，采用贪婪搜索策略，定义了紧密耦合CpG位点的区块，称之为甲基化单倍型区块（Methylation Haplotype Blocks, MHBs）。该数据集包含43组已公开的人类成人组织、胚胎干细胞（ESC, embryonic stem cell）及体外分化细胞系的WGBS数据，以及本研究中新生成的10组人类成人组织WGBS数据。随后，我们使用101组ENCODE项目的简化代表性亚硫酸氢盐测序（reduced representation bisulfite sequencing, RRBS）数据，以及637组来自癌症基因组图谱（The Cancer Genome Atlas, TCGA）肿瘤与正常样本的Illumina 450K甲基化芯片数据，对MHBs进行了进一步验证。在全基因组层面，MHBs在多个经典基因组特征区域中富集，但仅与这些区域部分重叠，包括CpG岛、启动子、增强子及可变甲基化区域（variable methylated regions, VMRs）。为实现对MHBs的定量分析，我们定义了一项名为甲基化单倍型负荷（Methylation Haplotype Load, MHL）的指标，该指标同时涵盖了平均甲基化水平与甲基化复杂度，因此相较于单一的平均甲基化水平或香农熵（Shannon entropy）具备更丰富的信息价值。通过特征选择策略，我们鉴定出一组按发育胚层聚类的组织特异性MHBs。有趣的是，对这些MHBs的分析揭示了发育过程中细胞命运决定的两种不同表观遗传机制：用于中胚层诱导的多能性基因（如NANOG）的表观遗传沉默，以及用于外胚层定向分化的谱系特异性因子的表观遗传诱导（或去抑制）。对实体瘤组织中的共甲基化模式进行分析