Bayesian Methods in Signal Transduction Network Analysis

描述（由申请人提供）：我们的长期目标是通过计算建模和整合不同类型的信息来研究细胞用于完成信号转导过程的机制。大规模基因组学和蛋白质组学技术的最新进展产生了大量的数据，并引起了我们对系统水平上信号转导的全面理解的注意。考虑到这些高通量技术产生的不完整和有噪声的数据，需要能够结合多个数据源来分析全基因组信号转导网络的新的计算方法，以充分利用数据的快速积累。在这项研究中，我们将开发强大的贝叶斯方法来整合不同的数据类型的信号网络推理，主要是在芽殖酵母酿酒酵母。我们提出的方法将被应用于鉴定蛋白质伴侣复合物和研究伴侣在介导信号通路中的作用。我们的方法的结果将受到进一步的实验验证。该应用程序的中心假设是，通过在异构大规模数据源上开发和应用贝叶斯方法，我们可以成功地推断全基因组范围内的信号网络。我们计划通过追求以下具体目标来验证假设并实现本申请的总体目标：1）开发基于大规模蛋白质相互作用数据的贝叶斯方法来识别蛋白质复合物; 2）通过整合来自不同来源的信息（包括微阵列基因表达数据、蛋白质相互作用数据和蛋白质磷酸化数据）来发现蛋白质之间的关联并推断信号转导网络; 3）开发方便用户的计算机软件来实现所提出的方法。该软件将免费开发、测试和分发给科学界。拟议的工作将代表通过数据集成从不同的大规模数据集中提取信息以推断全基因组范围内的信号转导网络的第一次重大努力。我们提出的方法可以成为一种强大的手段，使推断信号转导网络的过程更快，更容易，并产生指导实验设计的假设，从而导致更多的信息实验。该研究将有助于通过集成异构数据源设计高效的信令网络推理方法。这些方法和用户友好软件的开发将提供有用的工具，以更好地了解细胞如何响应环境变化，更重要的是，这些反应的失败如何导致各种疾病。

DESCRIPTION (provided by applicant): Our long-term goal is to investigate the mechanisms cells use to accomplish signal transduction processes through computational modeling and integration of different types of information. Recent advances in large- scale genomic and proteomic techniques have generated enormous amounts of data and have drawn our attention towards a comprehensive understanding of signal transduction at the systems level. Given the incomplete and noisy data generated from these high-throughput techniques, novel computational approaches capable of incorporating multiple data sources for analyzing the genome-wide signal transduction networks are needed to fully take advantage of the rapid accumulation of data. In this study, we will develop robust Bayesian methods to integrate diverse data types for signaling network inference, primarily in the budding yeast Saccharomyces cerevisiae. Our proposed methodology will be applied for identifying protein chaperone complexes and investigating the roles of chaperones in mediating signaling pathways. The results of our method will be subject to further experimental validation. The central hypothesis of the application is that, by developing and applying Bayesian methods on heterogeneous large-scale data sources, we can successfully infer signaling networks in a genome-wide scale. We plan to test the hypothesis and accomplish the overall objective of this application by pursuing the following specific aims: 1) Develop Bayesian methods to identify protein complexes based on large-scale protein interaction data; 2) Discover the associations among proteins and infer signal transduction networks by integrating information from diverse sources, including microarray gene expression data, protein interaction data and protein phosphorylation data; 3) Develop user-friendly computer software to implement the proposed methods. The software will be developed, tested and distributed to the scientific community free of charge. The proposed work will represent the first major effort that extracts information from diverse large-scale datasets through data integration for inferring signal transduction networks at a genome-wide scale. Our proposed approach can be a powerful means to make the process of inferring signal transduction networks faster and easier, and produce hypotheses that guide the experimental design, leading to more informative experiments. The research will contribute to designing efficient signaling network inference methods through integrating heterogeneous data sources. The development of these methods and user-friendly software will provide useful tools to better understand how cells respond to environment changes, and more importantly, how failure of these responses leads to a variety of diseases.