Advanced Computational Approaches to Delineating Dynamic Cancer Progression Processes by Using Massive Static Sample Data

摘要 人类癌症是一种动态的疾病，它在较长的一段时间内通过积累 一系列的基因改变。描绘疾病进展的系统动力学可以显著推进 我们对肿瘤生物学的了解，并为改善癌症的发展奠定了关键基础 诊断学、预见学和靶向治疗。传统上，系统动力学是通过 通过对同一队列的受试者在整个生物学中进行重复抽样而实现的时间进程研究 进程。然而，由于伦理和经济的限制，收集时间序列数据进行研究是不可行的 人类癌症，通常我们只能从切除的肿瘤组织中获得概况数据。因此，虽然 主要的努力继续揭示与人类癌症有关的基因组事件，到目前为止，这是困难的 将确定的变化放在动态疾病过程的背景下。随着……的快速发展 测序技术，成千上万的静态肿瘤样本正在大规模的癌症中收集 学习。这为我们提供了一个独特的机会来开发使用静态数据的新的分析策略， 而不是时间进程数据，来研究疾病动力学。基于我们以前的工作，我们提出了一种 大规模跨学科研究计划开发一系列新方法，使 利用海量静态数据建立高分辨率肿瘤进展模型，识别关键分子 推动疾病逐步发展的事件，以及癌症中已识别的变化的可视化 发展路线图。如果成功实施，这项工作可以有效地克服现有的采样 限制，并开辟了一条新的研究途径，通过使用大量的组织档案来研究癌症动力学 进行资源密集型或不切实际的时间课程研究。所开发的方法将集中于 在包含约9,000个样本的27个乳腺癌数据集上进行了测试。据我们所知，之前没有任何工作是 在这个量表上进行，以研究乳腺癌动力学。分析将产生第一个工作模式 通过整合所有遗传信息构建的乳腺癌进展。所构建的模型可以 为关键进展分子事件的可视化提供基础，并有助于识别 治疗干预的关键驱动基因和途径以及潜在的敏感点。此外， 对构建的模型的询问将使我们能够用Silico测试新的假设并确定优先顺序 更有针对性和更详细的试验性调查的资源。我们希望我们的工作将有一个 广泛的影响。虽然在这项研究中我们主要关注乳腺癌，但开发的方法也可以 用于研究其他癌症和其他人类进展性疾病，其中缺乏时间序列数据来研究 系统动力学是一个普遍存在的问题。

Abstract Human cancer is a dynamic disease that develops over an extended time period through the accumulation of a series of genetic alterations. Delineating the system dynamics of disease progression can significantly advance our understanding of tumor biology, and lay a critical foundation for the development of improved cancer diagnostics, prognostics and targeted therapeutics. Traditionally, system dynamics is approached through time-course studies achieved by repeated sampling of the same cohort of subjects across an entire biological process. However, due to ethical and economic constraints, it is not feasible to collect time-series data to study human cancer, and typically we can only obtain profile data from excised tumor tissues. Consequently, while major efforts continue to reveal the genomic events associated with human cancer, to date, it has been difficult to put the identified changes in the context of the dynamic disease process. With the rapid development of sequencing technology, many thousands of static tumor samples are being collected in large-scale cancer studies. This provides us with a unique opportunity to develop a novel analytical strategy to use static data, instead of time-course data, to study disease dynamics. Built logically on our previous work, we propose a large-scale interdisciplinary research plan to develop a series of novel methods that enable the construction of high-resolution cancer progression models by using massive static data, the identification of pivotal molecular events that drive stepwise disease progression, and the visualization of identified changes in a cancer development roadmap. If successfully implemented, this work can effectively overcome the existing sampling limitations, and open a new avenue of research to study cancer dynamics by using vast tissue archive, instead of performing resource-intensive or impractical time-course studies. The developed methods will be intensively tested on 27 breast cancer datasets comprised of ~9,000 samples. To our knowledge, no prior work has been performed on this scale to study breast cancer dynamics. The analysis will result in the first working model of breast cancer progression constructed by incorporating all genetic information. The constructed model can provide a foundation for the visualization of key progressive molecular events and facilitate the identification of pivotal driver genes and pathways and potential points of susceptibility for therapeutic intervention. Moreover, interrogation of the constructed model will enable us to test novel hypotheses in silico and to prioritize resources for more focused and detailed investigations experimentally. We expect that our work will have a broad impact. Although in this study we focus mainly on breast cancer, the developed methods can also be used to study other cancers and other human progressive diseases, where the lack of time-series data to study system dynamics is a ubiquitous problem.