Comprehensive functional characterization and dissection of noncoding regulatory elements and human genetic variation

摘要 ENCODE项目已经生成了全面的顺式调控元件(CRE)地图，控制着 人类基因组内基因的转录。这些地图在我们努力了解 与人类特征和疾病相关的序列变异，因为这些变异中的大多数都是非编码的 调节变化，而不是氨基酸替代。然而，即使我们知道 数以千计的CRE，我们对它们如何运行的理解来自于相对较小的一组 描述得很好的例子。因此，我们计划直接描述Encode Cres的功能 多种细胞类型的全基因组范围。这将使功能基因组学领域从简单的图谱转变为 对监管要素的深入理解，向下规范监管逻辑的根本规则 到Basepair解决方案。实现这一点将通过增强我们的能力来极大地扩展ENCODE的用途 解释人类自然变异对基因调控的影响。 我们建议直接测量超过3%的基因组的调节活性，寻找突出显示的基因座 ENCODE和其他函数数据很重要。我们将首先应用计算方法来确定大多数 生物信息性CRES，代表了各种监管逻辑和体系结构，并将使用 机器学习技术，用于确定与常见和罕见相关的特征的功能变体的优先顺序 人类疾病、特征和适应。我们将从中选择100,000个CRE和375,000个变种， 代表约100Mb的基因组序列，并使用大规模平行报告对其进行表征 测定(MPRA)以了解每个元件的调节活性。然后，为了补充来自MPRA的数据，我们 将使用基于CRISPR的非编码屏幕来表征20个基因座上的额外1 Mb区域，以构建 这些地点的全貌。该策略利用了MPRA的吞吐量和灵活性，同时 在基于CRISPR的屏幕中保持监管逻辑的连通性，这扰乱了 它们的内源基因组环境。这将有助于我们判断ENCODE的准确性和完整性，而 还提供了这两种方法的数据，以解决广泛的研究问题。这些方法是 很难全面应用于疾病相关的原代细胞，但我们将使用我们的MPRA和 CRISPR筛选为我们的模型提供信息，并更好地预测监管逻辑的基本规则。到时候我们会的 构建较小的、有针对性的文库，以测试原代细胞中的疾病特异性变体，并使用针对 三种自身免疫性疾病：1型糖尿病、炎症性肠病和狼疮。 这一方法将向研究界通报管理CRES活动的规则，由 ENCODE项目，并将同时提供有关数百个 数千个与人类特征、健康和疾病相关的序列变体。

Summary The ENCODE project has generated comprehensive maps of cis-regulatory elements (CREs) controlling the transcription of genes within the human genome. These maps have been crucial in our efforts to understand sequence variants linked to human traits and disease, as the majority of these variants are non-coding regulatory changes rather than amino acid substitutions. However, even though we know the locations of thousands of CREs, our understanding of how they operate is derived from a relatively small set of well-described examples. Therefore, we plan to directly characterize the function of ENCODE CREs at a genome-wide scale in multiple cell-types. This will transition the field of functional genomics from a simple map of regulatory elements towards a deep understanding of the fundamental rules governing regulatory logic down to the basepair resolution. Achieving this will dramatically expand ENCODE’s utility by strengthening our ability to interpret the effects of natural human variation on gene regulation. We propose to directly measure regulatory activity of over 3% of the genome, pursuing loci highlighted as important by ENCODE and other functional data. We will first apply computational methods to identify the most biologically informative CREs, representing a diversity of regulatory logic and architecture, and will use machine learning techniques to prioritize functional variants for characterization relevant to common and rare human diseases, traits, and adaptation. Of these we will select 100,000 CREs and 375,000 variants, representing ~100 Mb of genomic sequence, and characterize them using the massively parallel reporter assay (MPRA) to understand each element’s regulatory activity. Then, to complement data from the MPRA, we will characterize additional 1 Mb regions across 20 loci using CRISPR-based non-coding screens to build a comprehensive picture of these loci. This strategy leverages the throughput and flexibility of MPRA while maintaining the connectivity of regulatory logic in the CRISPR-based screens, which perturb elements within their endogenous genomic context. This will help us judge the accuracy and completeness of ENCODE, while also providing data from both approaches to address a wide-variety of research questions. These methods are difficult to apply to disease relevant primary cells at full scale, but we will use the results of our MPRA and CRISPR screens to inform our models and better predict the fundamental rules of regulatory logic. We will then construct smaller, targeted libraries to test disease-specific variants in primary cells and use assays specific for each of three autoimmune diseases: type 1 diabetes, inflammatory bowel disease, and lupus. This approach will inform the research community on the rules governing the activity of the CREs mapped by the ENCODE project, and will simultaneously provide concrete information about the function of hundreds of thousands of sequence variants relevant for human traits, health, and disease.