ITR: Feedback from Multi-Source Data Mining to Experimentation for Gene Network Discovery

在大约3万到3.5万个人类基因中，只有大约一半的基因功能是已知的。确定基因功能有许多不同的实验数据来源（例如序列、表达和蛋白质组学数据）。每个数据源的分析方法都有自己的优势，并且正在成熟，而这些现有算法的回报正在减少。为了扩大发现的范围，这项工作汇集了计算机科学和生物学领域的不同研究人员，以开发和应用数据挖掘方法，分析多种实验数据类型的多种来源。目标是发现人类和酵母基因的基因网络。这些方法能够识别和支持在孤立分析单一实验方法的数据时被忽视的生物学假设。人类和酵母基因的基因网络的发现有望解决诸如确定细胞中基因的基本组织和创建解释高通量生物学数据的理论框架等重大挑战问题，最终走向生物学的预测理论模型和在细胞水平上理解疾病。该项目还将帮助德克萨斯大学建立一个广泛的计算生物学卓越中心。除了通过项目网站{http://bioinformatics.icmb.utexas.edu}和科学出版物传播新算法外，新衍生的基因功能将提交给公共生物数据库，如BIND（生物分子相互作用网络数据库）和DIP（相互作用蛋白质数据库）。

Gene function is known for only about half of the roughly 30 to 35 thousand human genes. There are many diverse experimental sources of data for determining gene function (e.g. sequence, expression, and proteomic data). Analysis methods for each data source have their individual strengths and are maturing, while the returns from these existing algorithms are diminishing. To expand the scope of discovery this effort brings together diverse researchers within Computer Science and Biology in order to develop and apply data mining methods that analyze multiple sources of multiple experimental data types. The goal is to discover gene networks for human and yeast genes. These methods are able to identify and support biological hypothesis that are overlooked when data from a single experimental methodology is analyzed in isolation. Discovery of gene networks for human and yeast genes promises to address such grand challenge problems as determining the fundamental organization of genes in the cell and creating a theoretical framework for interpreting high-throughput biological data, moving ultimately towards predictive theoretical models of biology and understanding disease at the cellular level.The project will also help establish a broad center of excellence in computational biology at the University of Texas. In addition to the dissemination of new algorithms through the project Web site {http://bioinformatics.icmb.utexas.edu} and scientific publications, newly derived gene functions will be submitted to public biological databases suchas BIND (Biomolecular Interaction Network Database) and DIP (Database ofInteraction Proteins).