Multi-omic phenotyping of human transcriptional regulators

项目总结 细胞中空等位基因的分子表型(Moric)计划正在使用多种扰动策略来 实现NHGRI为每个人类基因赋予功能的愿景。策略包括集合策略和个人策略 基因敲除和敲除(KDS)，使用CRISPR技术和生长素诱导的退化产生。 在应用这种分析之后，细胞的分子表型被纵向地和在 使用散装和单细胞(Sc)RNA序列的单个时间点。扰动策略有内在的来源 可变性，例如KD外显率，而单细胞测序方法造成技术噪声，例如， “退学。”量化和控制这种变异性对于确保可靠的表型评估和 实现Moric的目标，准确地对基因功能进行分类。鉴于转录因子(TF)在转录因子中的关键作用 调节细胞状态，所有四个形态数据生产中心(DPC)都会扰乱TF，然后分析细胞 使用Bulk或sc-RNA-seq。广泛的其他“调节性表型”数据，包括(成批或单细胞) 在Moric和Tf CHIP-SEQ、HIC和大规模平行报告试验中产生ATAC-SEQ (MPRA)数据可在ENCODE和基因组变异对功能的影响(IGVF)联盟中获得。至 在稳健定义Tf扰动的调节影响的基础上，我们提出了JAX形态数据分析和验证 中心(DAV)分析这些多模式数据。我们的团队处于独特的地位，可以建立这个以TF为重点的 DAV：我们与JAX Moric DPC位于同一地点，并与其PI进行联盟级别的合作，而 我们自己的工作重点是阐明基因的转录调控和开发强大的计算 方法通过社区的努力。在目标1中，我们将量化和控制基于扰动的可变性 通过使用在Moric中生成的异类数据来隔离其技术噪声的调节性表型 并获得一组Tf基因靶点对(Tf-GTS)。然后我们将计算模拟大型- 比例扰动屏幕，我们将通过该屏幕执行功率分析，以量化数据可变性并做出 改善它的建议。在目标2中，我们将评估已发表的基因调控网络(GRN) 推理方法。我们还将进行两个“众包”梦想挑战，在其中的社区 参与者将开发GRN推理方法，我们将使用形态数据对其进行客观评估。使用TOP- 执行方法，以及我们正在开发的基于动力系统的新方法，我们将 在GRN中执行电子传递函数扰动，以确定用于在MOMIC中进行实验验证的传递函数的优先级。在AIM 3，我们将通过将推断的TF-GT与TF CHIP-SEQ、HIC和MPRA集成来进一步提高它们的健壮性 数据，敲除小鼠表型数据(KOMP2)，以及来自JAX和MOMIC的空间转录数据。我们 将验证已发表的定义组织特异性GRN的方法，方法是将其与相关形态重叠 模型系统，然后将使用它们来预测尚未由Moric分析的系统中的Tf-GT。我们的目标将 增强该领域破译人类基因组中~1600个Tf基因调控功能的能力。

PROJECT SUMMARY The Molecular Phenotypes of Null Alleles in Cells (MorPhiC) program is using multiple perturbation strategies to realize the NHGRI's vision of assigning function to every human gene. Strategies include pooled and individual gene knockouts and knockdowns (KDs), generated using CRISPR technologies and auxin-inducible degrons. Following application of such assays, molecular phenotypes of the cells are profiled longitudinally and at individual time points using bulk and single-cell (sc)RNA-seq. Perturbation strategies have intrinsic sources of variability, e.g., KD penetrance, while the single-cell sequencing approaches contribute technical noise, e.g., `drop out.' Quantifying and controlling this variability are crucial to ensure reliable phenotypic assessment and fulfill MorPhiC's goal to accurately catalog gene function. Given the critical role of transcription factors (TFs) in regulating cell state, all four MorPhiC Data Production Centers (DPCs) will perturb TFs and then profile cells using bulk or sc-RNA-seq. A wide range of other `regulatory phenotyping' data, including (bulk or single-cell) ATAC-seq, are being generated within MorPhiC and TF ChIP-seq, HiC, and massively parallel reporter assay (MPRA) data are available in the ENCODE and Impact of Genomic Variation on Function (IGVF) consortia. To robustly define the regulatory impact of TF perturbation, we propose a JAX MorPhiC Data Analysis and Validation Center (DAV) to analyze these multi-modal data. Our team is uniquely positioned to establish this TF-focused DAV: we are co-located with the JAX MorPhiC DPC and have consortium-level collaborations with its PI, while our own work focuses on elucidating transcriptional regulation of genes and on developing robust computational methods through community efforts. In Aim 1, we will quantify and control variability in perturbation-based regulatory phenotyping by using heterogeneous data generated within MorPhiC to isolate their technical noise characteristics and to derive a set of TF-gene target pairs (TF-GTs). We will then computationally simulate large- scale perturbation screens, through which we will perform power analysis to quantify data variability and make recommendations that ameliorate it. In Aim 2, we will evaluate published gene regulatory network (GRN) inference methods. We will also conduct two “crowd-sourced” DREAM Challenges, in which community participants will develop GRN inference methods that we will objectively evaluate with MorPhiC data. Using top- performing methods, as well as a novel approach we are developing based on dynamical systems, we will perform in silico TF perturbation within the GRNs to prioritize TFs for experimental validation in MorPhiC. In Aim 3, we will further improve robustness of inferred TF-GTs by integrating them with TF ChIP-seq, HiC, and MPRA data, knockout mouse phenotyping data (KOMP2), and spatial transcriptomics data from JAX and MorPhiC. We will validate published methods for defining tissue-specific GRNs by overlapping them with relevant MorPhiC model systems and will then use them to predict TF-GTs in systems yet to be profiled by MorPhiC. Our Aims will bolster the field's ability to decipher the regulatory function of the ~ 1600 TF genes within the human genome.