Core--Data Management and Computation Modeling

点击翻译按钮获取中文摘要

The goals of this Integrated Cancer Biology Program (ICBP) are to 1) increase our understanding of complex epigenetic alterations in neoplasms and 2) use this high-end information for improved prognosis, intervention and treatment of female cancers. In order to help accomplish these goals, the Core B investigators will : Specific Aim 1: Provide state-of-the-art computational and statistical support for maintaining and managing an interactive database. Using Java(TM) technology, we have developed Genome Data Visualization Toolkit (GDVTK) that consists of a set of data structures and core classes. This GDVTK is a sound framework for developing web-based applications to present the genomic annotations in visual form. We will employ GDVTK to develop a robust, flexible data management system for storage and query of promoter CpG islands and the associated methylation and genetic changes, histone modifications and chromatin status in cancer cell lines, neoplastic epithelium, and tumor stroma. Specific Aim 2: Develop innovative Bayesian methods to predict outcomes of epigenetic and genetic variables. Both supervised and unsupervised classification methods will be use for data mining of epigenomic results. Most of the machine-learning methods are data-intensive and susceptible to over-fitting, both of which lead to false-positive predictions when applied to new datasets. To address this concern, we will use a combination of cross-validation and permutation testing methods to produce robust statistical models. Specific Aim 3: Provide consultation in the analysis and reporting of microarray data produced in the proposed ICBP projects. For example, we will provide methods to address problems inherent in analyzing large, complex epigenomic data sets. This Core also integrates relevant data from ICBP projects with other distributed resources, such as GenBank and CaBIG (Cancer Biomedical Informatics Grid), into a centralized data warehouse. The database (http://bioinformatics.med.ohio-state.edu/ICBP) will be made available to all the investigators through a user-friendly web-interface.