Collaborative Research: Novel methods for pharmacogenomic data analysis using gene clusters

摘要大量的药物基因组学研究已经使用微阵列来调查整个基因组并检测与疾病相关的基因。基因具有固有的簇状结构。本研究的目的是开发一个系统的框架，使用基于主成分分析(PCA)的方法来检测差异表达和/或具有联合预测能力的基因簇。更具体地说，研究人员将(1)开发新的方法来检测略有差异表达的基因簇；(2)开发惩罚方法来检测对感兴趣的疾病临床结果具有联合预测能力的基因簇；以及(3)进行广泛的数值研究，并开发公开可用的软件。这项研究将极大地促进我们对？大p，小n？统计数据以及人类基因组学。本研究开发的方法学可应用于图像处理、免疫学、分子动力学、小角散射和信息检索等领域。从药物基因组数据分析中识别基因组标记是理解人类基因组学和个性化医学的关键一步。提出这项研究的动机是迫切需要克服现有方法的缺点。它将以新颖的统计方法、严谨的理论发展、广泛的数值研究、公共软件的开发以及对实践研究的直接影响为特色。拟议的研究总体上将丰富高维方法论家族。此外，对乳腺癌、结肠癌和淋巴瘤微阵列数据的分析将有助于更深入地理解这些癌症背后的基因组机制。从教育和社会的角度来看，这项拟议的研究将促进来自不同机构和背景的研究人员之间更紧密的合作。它将促进耶鲁大学和北卡罗来纳大学的教学、培训和学习。此外，研究人员将出席统计和基因组会议并发表演讲，这可能会促进来自不同领域的科学家之间的跨学科研究。

AbstractNumerous pharmacogenomic studies have been conducted using microarrays to survey the whole genome and detect disease-associated genes. Genes have the inherent clustering structure. The goal of this study is to develop a systematic framework using principal component analysis (PCA) based methods to detect gene clusters differentially expressed and/or with joint predictive power. More specifically, the investigators will (1) develop novel methodology to detect gene clusters marginally differentially expressed; (2) develop penalization methodology to detect gene clusters with joint predictive power for the disease clinical outcomes of interest; and (3) conduct extensive numerical studies, and develop publicly available software. This study will greatly advance our understanding of the ?large p, small n? statistics as well as human genomics. Methodologies developed in this study can be applied in other areas including image processing, immunology, molecular dynamics, small-angle scattering, and information retrieval.Identification of genomic markers from analysis of pharmacogenomic data is a key step in understanding human genomics and personalized medicine. The proposed study has been motivated by the urgent need to overcome drawbacks of existing methods. It will feature novel statistical methods, rigorous theoretical development, extensive numerical studies, development of public software, and a direct impact on practical studies. The proposed study will enrich the family of high dimensional methodologies in general. In addition, analysis of breast cancer, colon cancer, and lymphoma microarray data will lead to a deeper understanding of the genomic mechanisms underlying those cancers. From educational and social prospective, the proposed study will foster more intensive collaborations among investigators from different institutions and background. It will promote teaching, training and learning at Yale University and at the University of North Carolina. Moreover, the investigators will attend statistical and genomic conferences and give presentations, which may promote interdisciplinary research among scientists from diverse fields.