Integrating single-cell based transcriptomic signatures for identifying therapeutic targets of COPD

该提案的主要目标是应用新的机器学习和基于网络的方法来促进 发现疾病中的生物标志物和药物的治疗靶点。而单细胞RNA测序 （scRNAseq）技术已经能够在单细胞分辨率下进行基因表达谱分析，并有助于检测 许多不同细胞类型中数千个基因的扰动和潜在的新疾病机制， 从数千个候选物中鉴定可行的治疗靶点仍然是一个具有挑战性的问题。因此 对于确定一组高优先级和精简的潜在药物靶点至关重要， 可以在实验室里用有限的资源进行实验验证。这项提议的动机是最近 我们小组的发现：我们通过分析细胞类型特异性和疾病相关的变化， COPD和健康肺组织的单细胞转录组学谱。一个限制是，这些结果来自 来自最严重的COPD患者的子集，并且可能不会推广到所有COPD亚型。scRNA 由于噪声水平过高和稀疏性，研究在预测基因调控网络（GRN）方面也有局限性 的数据。因此，我们建议利用来自大型队列的批量转录组学数据的信息， 例如基因型-组织表达（GTEx）和肺基因组学研究联盟（LGRC）。我们 相信使用来自批量水平、大型队列、RNAseq数据的GRN作为COPD肺部的基线GRN将 导致更强大的识别疾病相关的细胞类型和途径，结果将更多 可推广至所有COPD亚型。在目标1中，我们将确定一系列转录因子（TF）， 在COPD中最活跃，并且最有可能调节我们的研究中鉴定的细胞类型特异性转录组特征。 scRNA研究。这将通过应用新的机器学习和网络方法来整合 来自GTEx和LGRC群组的GRN和scRNA数据。在目标2中，我们将应用这种方法来研究 吸烟（CS）。这一目标是由我们的scRNA-seq研究激发的，该研究鉴定了不同的基因， 两个AT 2亚群中的表达扰动。由于我们的COPD受试者是晚期患者， 这些结果可能反映了持续的病理变化， 戒烟后仍在继续。因此，我们将应用与目标1中相同的方法来识别细胞， COPD人肺组织中吸烟的类型特异性转录组特征和 很可能是这些签名的中介。在目标3中，我们将确定适合靶向TF的潜在药物列表 根据目标1和目标2的结果制定模块。我们将根据目标1和2将TF模块映射到 DrugBank是一个药物靶点发现数据库，我们将确定最有可能有效的药物， 基于网络与模块内TF的接近度来定位这些模块。我们认为这些药物和药物 这些靶点可能为我们提供最好的机会来验证新的TF/药物靶点，从而可能导致新的治疗策略。 对于COPD患者。

The main goal of this proposal is to apply novel machine learning and network-based methods to facilitate the discovery of biomarkers in diseases and therapeutic targets of drugs. While single-cell RNA sequencing (scRNAseq) technology has enabled gene expression profiling at single cell resolution, and has helped detect perturbation in thousands of genes across many different cell types and potential novel disease mechanisms, identifying viable therapeutic targets out of thousands of candidates is still a challenging problem. Therefore, it is essential to identify a high-priority and streamlined set of potential drug targets whose role in the disease could be experimentally validated in the lab with finite resources. This proposal is motivated by a recent discovery by our group: We identified both cell-type-specific and disease associated changes by analyzing single-cell transcriptomics profiles of COPD and healthy lung tissues. One limitation is that these results come from a subset of most severe COPD patients and may not be generalized to all COPD subtypes. The scRNA study also has limitation for predicting gene regulatory network (GRN), due to inflated noise level and sparsity of the data. Therefore, we propose to leverage information from bulk transcriptomic data from large cohorts, such as the Genotype-Tissue Expression (GTEx) and The Lung Genomics Research Consortium (LGRC). We believe that using GRN from bulk-level, large cohort, RNAseq data as the baseline GRN for COPD lungs will lead to more robust identification of disease-associated cell types and pathways, and the results will be more generalizable to all COPD subtypes. In Aim 1 we will identify a list of transcription factors (TFs) that that are most active in COPD and most likely to regulate the cell-type-specific transcriptomic signatures identified in our scRNA study. This will be achieved by applying novel machine learning and network methods to integrate the GRN from GTEx and LGRC cohorts and scRNA data. In Aim 2 we will apply this method to study the effects of cigarette smoking (CS). This aim is motivated by our scRNA-seq study which identified distinct gene expression perturbations in two AT2 subpopulations. As our COPD subjects were advanced patients who had stopped smoking in anticipation of their lung transplant, these results may reflect persistent pathologic changes that continue after smoking cessation. Therefore, we will apply the same approach as in Aim 1 to identify cell- type specific transcriptomic signatures of cigarette smoking in COPD human lung tissues and TFs that are likely to mediate these signatures. In Aim 3 we will identify a list of potential drugs suitable for targeting the TF modules based on the results from Aims 1 and 2. We will map the TF modules based on Aims 1 and 2 to DrugBank, a Drug Target Discovery database, and we will identify the drugs that are most likely to effectively target these modules based on network proximity to TFs within the module. We believe these drugs and drug targets may give us the best chance to validate novel TF/drug targets that may lead to new treatment strategy for COPD patients.