Engineering model-based systems to monitor and steer subclonal dynamics

摘要 原发性肿瘤以及癌细胞系已经显示出广泛的遗传和转录调控， 异质性，多个亚克隆共存于同一癌症群体中。1即使经过数十年的体外研究， 随着生长，已建立的细胞培养物继续发展。2癌细胞系在空间和时间上的异质性 具体为三个未满足的需求：i）提供体外生长的高时间分辨率的细胞培养方案 动力学; ii）密切监测基因型和表型测量之间的时间分离，以及 iii）将重复的高通量多组学测量的成本高昂的性质与不断的 亚克隆组成的变化。为了满足这些需求，我们建议工程师如何在体外和硅实验 与一个名为CLONEID的软件解决方案进行交互。后端的SOL数据库、Java核心和 R的用户界面将一起形成两个模块：一个将记录血统的血统在一个 实验室和使用计算机视觉来监测表型变化，如可变的生长率。第二模块 将来自不同高通量测定的亚克隆多组学图谱相互关联并与表型关联 从第一个模块。在aim 1中，我们将开发第一个模块，并使用它来证明监控的可行性 在没有任何专门设备的情况下，以高时间分辨率观察具有克隆ID的细胞系的表型转变。 Aim 2将使用这些数据与相同细胞系的现有单细胞测序相结合， 测试第二个模块，并使用它来识别生长的亚克隆特异性生物标志物。这些目标将 为更复杂的致癌作用数学模型铺平了道路，这些模型不必依赖于简化 假设单个驱动突变具有相等的适应度效应，并且单个亚克隆具有固定的 增速在未来，这里开发的框架将作为部署计算机视觉的基础， 早期检测形态学变化，包括从体内到体外生长的适应和支原体 污染.

ABSTRACT Primary tumors as well as cancer cell lines have been shown to exhibit extensive genetic and transcriptional heterogeneity, with multiple subclones co-existing in the same cancer population.1 Even after decades of in-vitro growth, established cell cultures continue to evolve.2 The heterogeneity of cancer cell lines over space and time crystallizes into three unmet needs: i) cell culture protocols that offer a high temporal resolution on in-vitro growth dynamics; ii) close monitoring of the temporal separation between genotypic and phenotypic measurements and iii) reconciling the cost-prohibitive nature of repeated high-throughput multi-omic measurements with ceaseless changes in subclonal composition. To fill these needs, we propose to engineer how in-vitro and in-silica experiments interact into a software solution called CLONEID. An SOL database in the backend, a Java core and an R user interface will come together to form two modules: One will record the pedigree of lineages grown in a lab and use computer vision to monitor phenotypic changes, such as variable growth rates. The second module will link subclonal multi-omics profiles from different high throughput assays to each other and to the phenotypes from the first module. In aim 1 we will develop the first module and use it to demonstrate feasibility of monitoring phenotypic transitions of cell lines with CLONE ID at high temporal resolution, without any specialized equipment. Aim 2 will use this data in conjunction with existing single cell sequencing of the same cell lines to develop and test the second module and use it to identify subclone-specific biomarkers of growth. Together these aims will pave the way to more complex mathematical models of carcinogenesis, that do not have to rely on the simplifying assumption that individual driver mutations have equal fitness effects and that individual subclones have a fixed growth rate. In the future, the framework developed here will serve as foundation to deploy computer vision for early detection of morphological changes, including adaptation from in-vivo to in-vitro growth and mycoplasma contamination.