Statistical Informatics for Cancer Research

描述(由申请人提供)：我们提出了一个计划项目，癌症研究中的统计信息学，以解决在基于人群的癌症研究中产生的高维数据分析所引发的一系列问题。该计划项目包括三个研究项目和两个核心。项目1的重点是对行政区域收集的疾病计数数据进行时空建模。具体目标的动机是在旨在监测和评估健康差距的流行病学研究中遇到的问题。我们提出的方法解决了与行政边界随时间变化、疾病计数稀少、空间混乱和大数据集的沉重计算负担相关的问题。方法将应用于来自三个州癌症登记的美国乳腺癌发病率数据、波士顿地区过早死亡率和NCI SEER数据。项目2的动机也是与癌症发病率和死亡率有关的空间索引数据，但重点是人口监测和空间集群检测。项目2的三个具体目标来自对NCI SEER数据的分析，另一个目标来自一项旨在评估儿童白血病空间聚集的病例/对照研究。该数据集还包括几个易感性遗传生物标记物的个体水平数据。该项目的一个子目标是通过研究疾病聚集模式是否因遗传多态而不同来评估基因-空间相互作用。项目3侧重于分析超高维基因组和蛋白质组生物标志物的方法。项目3也考虑了空间索引基因组数据的扩展。这三个项目的所有目标都与研究人员参与的激励性现实世界癌症研究紧密结合在一起。这三个项目通过关注以人群为基础的癌症观察性研究，以及从技术上考虑需要先进统计和计算方法的高维相关数据(来自不同来源)，将主题联系在一起。几种特定的技术(例如，时空建模、惩罚似然、错误发现率、隐马尔可夫模型)在两个项目之间共享，在某些情况下在所有三个项目之间共享。这两个核心由一个管理核心和一个统计计算核心组成。行政核心将协调整个计划的科学指导和规划活动，其中将包括短期课程、访客计划、研究成果的传播和外部咨询委员会。统计计算核心将确保开发和传播开放获取、高质量、用户友好的软件，以实施研究项目中开发的统计方法，这是三个项目的最终具体目标。项目主任和联席主任刘易斯·瑞安教授和林锡鸿教授分别是国际知名的生物统计学家，在学术管理方面有着良好的记录。

DESCRIPTION (provided by applicant): We propose a Program Project, Statistical Informatics in Cancer Research, to tackle a series of problems motivated by the analysis of high dimensional data arising in population-based studies of cancer. This Program Project comprises three research projects and two cores. Project 1 focuses on spatio-temporal modeling of disease count data collected for administrative areas. The specific aims are motivated by problems encountered in epidemiological studies designed to monitor and assess health disparities. Our proposed methods address issues associated with administrative boundaries changing over time, sparse disease counts, spatial confounding, and heavy computational burdens for large data sets. Methods will be applied to data on U.S. breast cancer incidence from three state cancer registries, Boston-area premature mortality, and NCI SEER data. Project 2 is also motivated by spatially-indexed data related to cancer incidence and mortality, but the emphasis is on population surveillance and spatial cluster detection. Three of the specific aims of Project 2 are motivated by the analysis of NCI SEER data and one from a case/control study designed to assess spatial clustering in childhood leukemia. This dataset also includes individual level data on several genetic biomarkers of susceptibility. One sub-aim of this project assesses gene-space interaction by studying whether disease clustering patterns differ according to genetic polymorphisms. Project 3 focuses on methods for the analysis of very high dimensional genomic and proteomic biomarkers. Extensions to spatially indexed genomic data are also considered in Project 3. All of the aims of the three projects are closely integrated with the motivating real world cancer studies in which the investigators are involved. The three projects link thematically through a focus on population-based, observational studies in cancer, as well as technically through the consideration of high-dimensional correlated data (arising from different sources) that require advanced statistical and computing methods. Several specific techniques (e.g. spatio-temporal modeling, penalized likelihoods, False Discovery Rates, hidden Markov models) are shared between two and in some cases all three projects. The two cores consist of an Administrative Core and a Statistical Computing Core. The Administrative Core will coordinate the overall scientific direction and programmatic activities of Program, which will include short courses, a visitor program, dissemination of research results, and an external advisory committee. A Statistical Computing Core will ensure the development and dissemination of open access, good quality, user friendly software designed to implement the statistical methods developed in the Research Projects, which is the final Specific Aim of each of the three projects. The Program Director and Co-Director, Professors Louise Ryan and Xihong Lin, respectively, are internationally known biostatisticians with strong track records of academic administration.