Statistical Model Building for High Dimensional Biomedical Data

描述(申请人提供)：当前大规模生物医学数据的典型特征是观察到的样本数量较少，以及广泛观察到的样本异质性。在微阵列数据分析中，识别与样本表型相关的差异表达基因(例如，癌症疾病发展)并根据基因表达预测样本表型是一些核心研究问题。现有的大多数统计方法都忽略了样本的异质性，从而失去了力量。 该项目建议开发新的统计方法，明确解决小样本量和样本异质性问题，并且可以非常普遍地应用。这些方法的有效性将通过来自肺和肾移植研究项目的大规模生物医学数据来展示。这些移植项目旨在通过识别分子生物标志物来预测同种异体移植排斥反应，从而提高肺/肾移植排斥反应的分子诊断和治疗水平，以便进行关键的早期治疗和快速、非侵入性和经济的检测。 具体目标是1)开发新的统计方法用于差异基因表达检测，明确地对样本异质性进行建模。2)发展新的统计方法，对高维生物医学数据进行分类，并纳入样本异质性。3)开发用于联合分析一组基因(例如，路径中的基因)的新的统计方法。4)使用开发的模型和方法回答肺和肾移植项目中与公共卫生相关的研究问题；并在用户友好和有良好文档记录的软件中实施和验证所提出的方法，并免费分发给科学界。 识别肺和肾移植受者排斥反应的新生物标志物是非常重要的。快速、可靠地检测和预测容易获得的体液中的排斥反应，可能会使临床干预试验迅速推进。我们建议研究分析大规模生物医学数据的新方法，以充分发挥其在关键早期治疗的移植排斥反应预测和分子诊断方面的潜力。

DESCRIPTION (provided by applicant): Typical of current large-scale biomedical data is the feature of small number of observed samples and the widely observed sample heterogeneity. Identifying differentially expressed genes related to the sample phenotye (e.g., cancer disease development) and predicting sample phenotype based on the gene expressions are some central research questions in the microarray data analysis. Most existing statistical methods have ignored sample heterogeneity and thus loss power. This project proposes to develop novel statistical methods that explicitly address the small sample size and sampe heterogeneity issues, and can be applied very generally. The usefulness of these methods will be shown with the large-scale biomedical data originating from the lung and kidney transplant research projects. The transplant projects aimed to improve the molecular diagnosis and therapy of lung/kidney allograft rejection by identifying molecular biomarkers to predict the allograft rejection for critical early treatment and rapid, noninvasive, and economical testing. The specific aims are 1) Develop novel statistical methods for differential gene expression detection that explicitly model sample heterogeneity. 2) Develop novel statistical methods for classifying high-dimensional biomedical data and incorporating sample heterogeneity. 3) Develop novel statistical methods for jointly analyzing a set of genes (e.g., genes in a pathway). 4) Use the developed models and methods to answer research questions relevant to public health in the lung and kidney transplant projects; and implement and validate the proposed methods in user-friendly and well-documented software, and distribute them to the scientific community at no charge. It is very important to identify new biomarkers of allograft rejection in lung and kidney transplant recipients. The rapid and reliable detection and prediction of rejection in easily obtainable body fluids may allow the rapid advancement of clinical interventional trials. We propose to study novel methods for analyzing the large-scale biomedical data to realize their full potential of molecular diagnosis and prognosis of transplant rejection prediction for critical early treatment.