A data-driven bioinformatics platform for the design and analysis of multiplexed antibody-based cytometry experiments in cancer research

摘要 肿瘤的进展、对治疗的抵抗和转移与肿瘤细胞生态系统的特征密切相关。基于抗体的多重细胞术是组织组成、发病机制和免疫浸润的表型表征的标准方法，具有单细胞（有时是空间）分辨率。这些数据中的细胞群的识别通过根据其抗原谱聚类细胞的算法以及通过历史上通过反复试验进化的预定义的标记物集来促进。然而，这些数据的注释是一个手动、主观和费力的过程，这阻碍了结果的可重复性和准确性。设计包括组织中存在的所有细胞类型和状态的特异性标志物的抗体组通常是不可行的，并且常用标志物的效率是未知的。因此，细胞簇可以在它们的抗原谱上几乎没有区别，或者包含细胞类型的混合物。为了克服这些限制，该项目将开发信息技术，利用现有的单细胞转录组图谱，以帮助和自动化的设计和分析的多重抗体为基础的细胞计数实验。我们的工作假设是，大量可用的组织单细胞转录组数据可以为细胞计数实验的设计、注释和分析提供信息。我们将开发和评估信息学技术，用于建立参考抗原谱和基于单细胞蛋白质转录组学数据的最佳抗体组（目标1），以及用于在基于抗体的多重细胞计数实验中自动识别，注释和门控细胞群（目标2）。这些新的计算方法将使任何研究人员能够1）基于储存库中托管的参考单细胞数据自动识别和注释细胞计数数据集中的细胞群，2）定义用于分选细胞群的最佳门，3）跨实验转移门，4）为给定组织或一组细胞群设计最佳抗体组，以及5）推断细胞的基因表达谱。我们将在一个开源软件和在线门户网站中实施这些方法，用于转录组指导的肿瘤细胞计数数据的注释和分析，并将通过几个计划的研讨会和教程与最终用户密切合作，以最大限度地提高该平台的实用性和推广范围（目标3）。我们将在白血病和胰腺癌组织上测试我们的平台，并使用光谱流式细胞术和多重定量免疫组织化学进行分析。该项目开发的信息学技术将通过提高肿瘤组织的多重抗体细胞术分析的表型分辨率、准确性和重现性来改变癌症研究。

ABSTRACT Tumor progression, resistance to therapy, and metastasis are closely related to the characteristics of the tumor cell ecosystem. Multiplexed antibody-based cytometry is the standard method for phenotypic characterization of tissue composition, pathogenesis, and immune infiltration with single-cell (and sometimes spatial) resolution. The identification of cell populations in these data is facilitated by algorithms that cluster cells according to their antigenic profile, as well as by predefined sets of markers that have historically evolved by trial and error. However, the annotation of these data is a manual, subjective, and laborious process that hinders the reproducibility and accuracy of the results. The design of antibody panels that include specific markers for all cell types and states present in a tissue is usually unfeasible, and the efficiency of commonly used markers is unknown. Consequently, cell clusters can differ little in their antigenic profile or contain a mixture of cell types. To overcome these limitations, this project will develop informatics technologies that leverage existing single-cell transcriptomic atlases to assist and automate the design and analyses of multiplexed antibody-based cytometry experiments. Our working hypothesis is that the vast amount of available single-cell transcriptomic data of tissues can inform the design, annotation, and analysis of cytometry experiments. We will develop and evaluate informatics technologies for establishing reference antigenic profiles and optimal antibody panels based on single-cell proteotranscriptomic data (Aim 1 ), and for automating the identification, annotation, and gating of cell populations in multiplexed antibody-based cytometry experiments (Aim 2). These new computational methods will enable any researcher to 1) automatically identify and annotate cell populations in a cytometry dataset based on reference single-cell data hosted in a repository, 2) define optimal gates for sorting cell populations, 3) transfer gates across experiments, 4) design optimal antibody panels for a given tissue or set of cell populations, and 5) infer the gene expression profile of cells. We will implement these methods in an open-source software and online portal for the transcriptome-guided annotation and analysis of cytometry data of tumors, and will closely work with end-users through several planned workshops and tutorials to maximize the utility and outreach of this platform (Aim 3). We will test our platform on leukemic and pancreatic cancer tissues profiled with spectral flow cytometry and multiplexed quantitative immunohistochemistry. The informatics technologies developed in this project will transform cancer research by boosting the phenotypic resolution, accuracy, and reproducibility of multiplexed antibody-based cytometry analyses of tumor tissues.