Studying transcription factor binding networks in activated and dying macrophages

背景：基因表达和细胞反应的调控是通过转录因子（tf）的结合在DNA上进行协调的。TF结合是一个受细胞外信号修饰的动态特异性过程。然而，人们对tf控制基因表达的复杂相互作用知之甚少。在这里，我们将专注于在下一代测序数据集上使用机器学习来解释骨髓源性巨噬细胞（bmdm）中的TF结合模式，并研究细胞环境的变化如何改变TF结合。目的A：研究Nf-？BMDMs中的TF绑定网络。在这里，我将使用ChIP-seq和RNA-seq在LPS激活的bmdm中定义指定Nf-？家族成员绑定。然后，我将使用CRISPR-Cas9在全球范围内测试我们的模型，选择性地去除预测会选择Nf-？家庭成员入住。然后，我将使用注意力神经网络和弹性网络回归来建立这个模型，以定义特定于每个Nf-？家族成员和他们控制的基因组。我还将通过CRISPR-Cas9在特定位点删除基序来更具体地测试我们的模型。目的B：利用细胞因子激活BMDMs研究TF结合动力学。为了研究TF结合的上下文变化，我们将用一组细胞因子治疗BMDMs，并进行ATAC和RNA测序。我们将使用Aim 1中开发的分析工具来定义每种情况下的TF结合模式以及与基因相关的TF组合。由相同TF或TF组合控制的基因组将进行通路分析，以确定每个TF和TF组的作用。我们还将研究不同细胞因子处理引起的细胞环境变化如何通过TF靶向的重排改变基因集。利用这些数据，我们将建立基因和TF集，以确定细胞因子在bmdm中的治疗。然后，我们将使用CRISPR/Cas9敲除在细胞因子背景下的基因集中被预测为重要的tf，以确认我们的结果并构建特定于环境的人工启动子。目的C：使用TF网络在单细胞RNA-seq数据中定义活/死BMDM细胞群。在这里，我们将诱导小鼠bmdm细胞凋亡、坏死和焦亡，并进行单细胞多组测序（ATAC和RNA）。将提取RNA表达和TF结合位点，并根据假时间对细胞进行排列，以确定细胞死亡的进展。然后，我们将像在目标2中一样，对每个细胞使用elasticnet回归来定义每种类型细胞死亡的转录调控网络。然后，我们将使用这些数据生成细胞死亡图谱，并在先前发表的单细胞实验中预测细胞死亡的数量和类型，这些实验已得到显微镜确认。这将使我们能够在单个细胞数据集中分离死细胞，并在实验中确定细胞死亡的类型。

Background: Control of gene expression and cell responses is coordinated on the DNA through the binding of transcription factors (TFs). TF binding is a dynamic and specific process that is modified by extracellular signals. However, the complex interplay of TFs to control gene expression is poorly understood. Here, we will focus on the use of machine learning on next generation sequencing datasets to interpret TF binding patterns in bone marrow derived macrophages (BMDMs) and study how changes in the cellular environment modify TF binding. Objective A: Studying the nuances in Nf-?B TF binding networks in BMDMs. Here, I will use ChIP-seq and RNA-seq in BMDMs activated with LPS to define the TFs which specifies Nf-?ß family member binding. I will then test our model globally by using CRISPR-Cas9 to selectively remove TFs which are predicted to select Nf-?ß family members for occupancy. I will then build upon this model using an attention neural network and elasticnet regression to define the transcriptional factor motif sets specific to each Nf-?ß family members and the gene sets they control. I will also test our model more specifically by CRISPR-Cas9 deletion of motifs at specific sites. Objective B: Studying the dynamics of TF binding using cytokine activation of BMDMs. To study contextual changes in TF binding, we will treat BMDMs with a panel of cytokines and perform ATAC and RNA sequencing. We will use the analysis tools developed in Aim 1 to define the TF binding patterns in each context and TF combinations associated with genes. Gene groups which are controlled by the same TF or combination of TFs will undergo pathways analysis to define the role of each TF and set of TFs. We will also study how changes in cellular context caused by different cytokine treatments change the gene sets through a rearrangement of TF targeting. Using this data, we will build gene and TF sets which define cytokine treatments in BMDMs. We will then use CRISPR/Cas9 to knockout TFs which are predicted to be important in gene sets in cytokine contexts to confirm our results and build context specific artificial promoters. Objective C: Defining live/dead BMDM cell populations in single cell RNA-seq data using TF networks. Here, we will induce apoptosis, necrosis and pyroptosis in mouse BMDMs and perform single-cell multi-omic sequencing (ATAC and RNA). RNA expression and TF binding sites will be extracted and cells will be aligned based on pseudotime to define the progression of cell death. We will then use elasticnet regression as in Aim 2 for each cell to define transcriptional regulatory networks for each type of cell death. We will then use this data to produce a cell death atlas and predict the amount and type of cell death in previously published single cell experiments which have microscopy confirmation. This will allow us to isolate dead cells in single cell datasets and determine the type of cell death in experiments.