Statistical Models in Epigenomics

描述(申请人提供)：我们的主要目标是利用“分子时钟假说”来开发数学模型，使我们能够研究癌症是如何生长和扩散的。人类癌症的生长不能直接观察，总体目标是开发一种方法，可以通过“读取”复制错误秘密写入基因组中的祖先来追溯重建肿瘤的进展。序列通常用于重建物种和个体的系谱，我们建议将这种一般的分子系统学方法翻译到人类癌症上。我们将使用DNA甲基化数据，这是一种在细胞分裂时复制的DNA的表观遗传修饰。由于直接计算可能不切实际或不可行，我们建议使用拒绝算法，这是一种基于模拟的方法。这一总体框架将使我们能够估计肿瘤的年龄、转移的年龄、甲基化错误率，以及转移是否来自原发癌的特定细胞群。我们的目标是受到南加州大学诺里斯综合癌症中心正在进行的研究的推动。具体地说，我们建议：1.开发方法，使我们能够使用5‘到3’DNA甲基化模式估计表征癌症生长的参数，并使用来自患者的临床数据和来自癌细胞系的实验数据来验证这些模型。这些模型将解决以下生物学问题：a.根据从甲基化模式推断的祖先树的类型估计癌症干细胞的数量；b.评估肿瘤的异质性，例如识别肿瘤左侧和右侧的不同细胞亚群；c.估计肿瘤的年龄和甲基化错误发生的比率。2.扩展目标1中开发的模型，以包括额外的复杂性。我们建议解决以下问题：a.对单个祖先树中的多个基因区域进行建模；b.对常染色体基因(二倍体基因组)进行建模；c.对多个组织(原发肿瘤和转移)进行建模，以回答有关细胞群体是否年龄相同或其中一个较年轻且源自另一个细胞群体的问题；我们将应用AIMS 1-2中开发的方法来研究在人类原发肿瘤和远处转移中观察到的DNA甲基化模式。公共卫生相关性：根据国家卫生统计中心的报告，癌症是2005年美国第二大死亡原因。它的治疗依赖于了解癌症是如何生长和扩散的。我们建议开发可以追溯重建肿瘤历史的数学模型，使我们能够解决关于癌症生长和扩散的重要生物学问题。

DESCRIPTION (provided by applicant): Our primary objective is to utilize the "molecular clock hypothesis" to develop mathematical models that will allow us to study how cancers grow and spread. Human cancer growth cannot be directly observed and the overall goal is to develop an approach that can retrospectively reconstruct tumor progression by "reading" the ancestry surreptitiously written within genomes by replication errors. Sequences are commonly used to reconstruct the genealogy of species and individuals, and we propose to translate this general molecular phylogeny approach to human cancers. We will use DNA methylation data, an epigenetic modification of DNA that is replicated at cell division. As direct calculation can be either impractical or infeasible, we propose to use rejection algorithms, a simulation-based approach. This general framework will allow us to estimate the age of a tumor, the age of a metastasis, the methylation error rate, and whether the metastasis is derived from a specific population of cells from the primary cancer. Our aims are motivated by ongoing studies at the Norris Comprehensive Cancer Center at the University of Southern California. Specifically, we propose to: 1. Develop methods that will allow us to estimate parameters characterizing the growth of cancer using 5' to 3' DNA methylation patterns and validate these models using clinical data from patients and experimental data from cancer cell lines. The models will address the following biological problems: a. Estimate the number of cancer stem cells based on the types of ancestral trees inferred from the methylation patterns; b. Evaluate tumor heterogeneity, e.g. identify different subpopulations of cells in the left and right side of the tumor; c. Estimate tumor age and the rate at which methylation errors occur. 2. Extend the models developed in Aim 1 to include additional complexities. We propose to address the following: a. Modeling multiple gene regions within a single ancestral tree; b. Modeling autosomal genes (diploid genomes); c. Modeling multiple tissues (primary tumor and metastasis), to answer questions about whether the cell populations are the same age, or if one is younger and derived from the other; We will apply the methods developed in Aims 1-2 to DNA methylation patterns observed in primary tumors of the colon and distant metastasis in humans. PUBLIC HEALTH RELEVANCE: Cancer is the second leading cause of death in the United States in 2005, as reported by the National Center for Health Statistics. Its treatment relies on understanding how cancers grow and spread. We propose to develop mathematical models that can retrospectively reconstruct tumor histories, allowing us to address important biological questions about the growth and spread of cancer.