Integrative and multiscale analysis of epigenomic sequencing data

“表观基因组”由大量的化学指令组成，用于从DNA中构建蛋白质的过程。这些指令控制着基因如何以及何时开启和关闭。不同的细胞类型、发育阶段和环境暴露都有各自的一套指令。解码这些指令对于理解基因调控及其变异的基本分子过程至关重要。虽然新技术现在可以在更多的维度和更高的分辨率上探测表观基因组，但我们解释这些测量结果的能力远远落后于我们生成它们的能力。从表观基因组实验中提取有意义的生物学信息的瓶颈在于数据分析和解释。我的研究项目侧重于开发和应用量身定制的统计方法来表征表观遗传变异及其对基因表达的影响。特别是，一个关键的差距在于难以整合来自不同样本和测量来源的表观基因组信息。这种挑战的部分原因是存在不受控制的技术工件，这些工件可以掩盖生物差异。另一个主要障碍源于这样一个事实，即表观遗传变化可以发生在许多不同的尺度上，涉及从单个核苷酸到数百万个核苷酸的任何地方，但现有的分析方法只能一次检查一个尺度。为了解决这些缺点，我的研究目标集中在开发新的量化框架，用于规范化、多尺度推理和表观基因组测序数据的综合建模。具体来说，我们将开发一种统计方法，用于计算校正最近发现的甲基化数据中可能混淆生物信号解释的测序偏差。为了在多个分辨率尺度上推断表观遗传变化，我们将开发一种从数据中推断尺度的统计方法。此外，我们将对表观遗传修饰与空间关系基因表达之间的关系进行系统的计算机评估。该研究促进了表观基因组信息的整合，这将有助于更好地理解基因调控的分子过程及其与环境或年龄等因素相关的变异。分析程序的关键进展将允许无偏解释，这也是信号占主导地位的尺度的不可知论。系统地探索基因间和基因内修饰的调控潜力将使我们深入了解表观基因组信号如何相互作用。此外，开发的方法将通过公开可用的工具传播，这将使社区中的其他研究人员能够从测序数据中提取有意义的生物学见解。

The `epigenome' consists of a multitude of chemical instructions for the process of building proteins from DNA. These instructions govern how and when genes are turned on and off. Different cell types, developmental stages, and environmental exposures can each have their own set of instructions. Decoding these instructions is vital to understanding the fundamental molecular processes of gene regulation and their variation. While new technologies can now probe the epigenome in more dimensions and at higher resolution than ever before, our ability to interpret these measurements lags far behind our ability to generate them. The bottleneck in extracting meaningful biological information from epigenomic experiments lies in data analysis and interpretation. My research program focuses on developing and applying tailored statistical methodology to characterize epigenetic variation and its effects on gene expression. In particular, a critical gap lies in the difficulty in integrating epigenomic information from different sample and measurement sources. Part of this challenge results from the presence of uncontrolled technical artifacts that can mask biological differences. Another major obstacle stems from the fact that epigenetic changes can take place at many different scales, involving anywhere from a single nucleotide to millions, but existing analytical approaches are hardwired to examine one scale at a time. To address these shortcomings, my research objectives focus on developing new quantitative frameworks for normalization, multiscale inference, and integrative modeling of epigenomic sequencing data. Specifically, we will develop a statistical approach for the computational correction of recently identified sequencing biases in methylation data that can confound the interpretation of biological signals. To infer epigenetic changes at multiple scales of resolution, we will develop a statistical approach that infers the scale from the data. In addition, we will carry out a systematic in silico evaluation of the relationship between epigenetic modification and gene expression that accounts for spatial relationships. The proposed research facilitates the integration of epigenomic information, which will lead to greater understanding of the molecular processes governing gene regulation and their variation with respect to factors such as environment or age. Key advances in analytic procedures will allow for unbiased interpretation that is also agnostic to the scale at which the signal predominates. Systematic exploration of the regulatory potential of inter- and intra-genetic modifications will yield insight into how epigenomic signals interact. Further, the methodology developed will be disseminated through publicly available tools, which will enable other investigators in the community to extract meaningful biological insights from sequencing data.