Improved Pattern Recognition for Functional Genomics

描述（由申请人提供）：计算方法已成为生物医学研究的固有内容。总体目标是为杨嘉仪博士（Ph.D.在计算机科学）与指导的培训和研究经验，过渡到一个独立的多学科的生物医学研究调查。为此目的设计了一个指导研究，学术课程和研究计划的方案。指导研究部分由导师和一个咨询委员会组成，他们是分子生物学、蛋白质组学、医学研究、生物信息学和统计学方面的顶尖专家。学术课程部分将为杨博士提供分子生物学，癌症生物学和统计学的坚实背景。研究计划的基本主题是开发方法和软件工具，以促进从癌症和疾病调查的高通量数据中提取生物学意义。 我们的研究计划的主要目标如下：具体目标1：开发改进的算法，用于肝细胞癌（HCC）和丙型肝炎病毒（HCV）相关肝脏疾病的微阵列数据的类预测和基因标记的识别。预测给定组织样本的诊断或预后类别（类别预测）和从微阵列数据中识别潜在基因标记的问题已经受到了很多关注。我们将利用微阵列数据中重复测量的变异性来开发改进的算法，用于类别预测和潜在基因标记的识别。具体目标2：在异构数据上开发类预测和类发现算法。我们将建立在我们以前的聚类分析和类预测的工作，开发算法来处理来自多个来源的数据，包括微阵列数据，蛋白质组学数据和临床数据。具体目标3：开发改进的可视化工具。将开发可视化软件工具，以方便生物学家利用他们的生物学知识，并解释高通量数据的计算结果。具体目标4：制定微阵列数据聚类分析的实用指南。我们将利用我们内部的数据库，包括数千个微阵列实验进行实证研究，以制定实用的聚类分析指南。

DESCRIPTION (provided by applicant): Computational methods have become intrinsic to biomedical research. The overall goal is to provide Dr. Ka Yee Yeung (Ph.D. in Computer Science) with mentored training and research experience to transition into an independent multi-disciplinary investigator in biomedical research. A program of mentored research, academic coursework, and research plan has been designed for this purpose. The mentored research component consists of mentors and an advisory committee who are leading experts in molecular biology, proteomics, medical research, bioinformatics and statistics. The academic coursework component will provide Dr. Yeung with a solid background in molecular biology, cancer biology and statistics. The underlying theme of the research plan is development of methods and software tools to facilitate extraction of biological meanings from high throughput data in cancer and disease investigation. The major goals of our research plan are the following: Specific Aim 1: Development of improved algorithms for class prediction and identification of gene markers on microarray data related to Hepatocellular carcinoma (HCC) and Hepatitis C virus (HCV) associated liver disease. The problems of predicting the diagnostic or prognostic category of a given tissue sample (class prediction) and identifying potential gene markers from microarray data have received a lot of attention. We will develop improved algorithms for class prediction and identification of potential gene markers by taking advantage of variability over repeated measurements in microarray data. Specific Aim 2: Development of class prediction and class discovery algorithms on heterogeneous data. We will build on our previous work in cluster analysis and class prediction to develop algorithms to handle data from multiple sources, including microarray data, proteomics data and clinical data. Specific Aim 3: Development of improved visualization tools. Software tools for visualization will be developed to facilitate biologists to utilize their biological knowledge and to interpret computational results from high throughput data. Specific Aim 4: Development of practical guidelines for cluster analysis on microarray data. We will make use of our in-house database consisting of thousands of microarray experiments to conduct empirical studies to develop practical guidelines for cluster analysis.