Novel Methods for Integrative Analysis of Cancer Genomic Data

描述（由申请人提供）：已使用高通量分析技术广泛进行了癌症基因组研究。从这些研究中鉴定的分子特征已用于辅助临床实践，包括诊断、预后预测和治疗方案的选择。尽管取得了令人鼓舞的成功，但这些签名往往缺乏可重复性和可靠性。这个问题的一个主要原因是样本量相对较小，因此缺乏个别研究的力量。一个具有成本效益的补救措施是汇集和分析来自多项研究的数据。用于分析多个数据集的可用方法具有严重的缺点。迫切需要新的统计方法，可以有效地分析和提取有用的信息，从多个癌症基因组研究。该项目将是系统地制定和实施综合分析方法的首批项目之一。所提出的方法将能够有效地分析来自多个癌症基因组研究的异质高维数据集。他们将能够解释多个基因组测量和通路结构在建模癌症发展中的联合作用，并能够适当调整临床和环境风险因素。通过开发R包和公共网站进行传播，将使我们的研究可供一般生物医学界使用。对多种癌症临床结果的数据分析将导致识别临床上有用的标志物。具体而言，我们计划（1）开发用于多个异质性癌症基因组数据集的整合分析的惩罚边缘筛选方法;（2）开发用于多个异质性癌症基因组数据集的整合分析的基于个体标记的惩罚方法;（3）开发用于多个异质性癌症基因组数据集的整合分析的基于途径的惩罚方法;（4）开发能够适当适应部分线性临床和环境协变量影响的综合分析方法;（5）传播所提出的方法，分析多种癌症的数据，并识别癌症标志物。拟议的研究将同样强调开发新的方法及其实际应用。它将为多个异质数据集的综合分析方法做出重大贡献，并使研究人员能够更有效地从癌症基因组研究中提取有用的信息。 公共卫生相关性：这项研究将是第一个系统地开发和实施新的综合分析方法，可以有效地分析多个异质性和高维癌症基因组研究。它将丰富综合分析的方法，使研究人员能够更有效地从现有数据中提取有用的信息，并导致更好地了解癌症基因组学。所提出的方法的应用将导致临床上有用的癌症标志物的鉴定。

DESCRIPTION (provided by applicant): Cancer genomic studies have been extensively conducted using high-throughput profiling techniques. Molecular signatures identified from these studies have been used to assist clinical practice including diagnosis, prognosis prediction, and selection of treatment regimens. Despite promising successes, these signatures often suffer from a lack of reproducibility and reliability. A major cause of this problem is the relatively small sample sizes and hence lack of power of individual studies. A cost-effective remedy is to pool and analyze data from multiple studies. Available methods for analyzing multiple datasets have serious drawbacks. There is an urgent need for novel statistical methodologies that can effectively analyze and extract useful information from multiple cancer genomic studies. This project will be among the first to systematically develop and implement integrative analysis methodologies. The proposed methods will be able to effectively analyze heterogeneous high-dimensional datasets from multiple cancer genomic studies. They will be able to account for the joint effects of multiple genomic measurements and the pathway structure in modeling cancer development, and be able to properly adjust for clinical and environmental risk factors. Dissemination through the development of R package and public website will make our research accessible to the general biomedical community. Analysis of data on multiple cancer clinical outcomes will lead to identification of clinically useful markers. Specifically, we plan to (1) Develop penalized marginal screening methods for integrative analysis of multiple heterogeneous cancer genomic datasets; (2) Develop individual-marker based penalization methods for integrative analysis of multiple heterogeneous cancer genomic datasets; (3) Develop pathway based penalization methods for integrative analysis of multiple heterogeneous cancer genomic datasets; (4) Develop integrative analysis methods that can properly accommodate partially linear clinical and environmental covariate effects; (5) Disseminate the proposed methods, analyze data on multiple cancers, and identify cancer markers. The proposed study will emphasize equally development of novel methodologies and their practical applications. It will make significant contributions to methodologies for integrative analysis of multiple heterogeneous datasets, and enable researchers to more efficiently extract useful information from cancer genomic studies. PUBLIC HEALTH RELEVANCE: This study will be among the first to systematically develop and implement novel integrative analysis methods, which can effectively analyze multiple heterogeneous and high-dimensional cancer genomic studies. It will enrich the family of methodologies for integrative analysis, enable researchers to more efficiently extract useful information from existing data, and lead to a better understanding of cancer genomics. Applications of the proposed methods will lead to identification of clinically useful cancer markers.