EDAC: ENCODE Data Analysis Center

描述（由申请人提供）：ENCODE数据分析中心（EDAC）提案旨在为ENCODE项目提供灵活的分析资源。ENCODE计划是一个大型的多中心计划，旨在定义人类基因组中的所有功能元素。这将使用许多不同的实验技术加上许多计算技术来实现。交付这组功能元素的关键部分是来自多个数据源的数据集成。ED AC提案旨在提供这种整合。根据RFA对该提案的禁止，EDAC工作的精确优先级将由外部小组分析工作组（AWG）设定。根据以往的经验，这些分析方法将需要各种各样的技术。我们期望必须应用复杂的统计模型来整合数据，特别是减轻人类基因组变量的广泛异质性和相关性的问题。我们有统计专家，他们可以利用人类基因组的庞大规模，加上有限数量的合理假设，来产生统计技术，这些技术对这种相当大的异质性是稳健的。我们还希望应用机器学习技术来构建组合数据集的集成方法。其中包括基于贝叶斯的推理方法和支持向量机的鲁棒计算机科学技术。这些方法都在ENCODE试点项目中表现良好，我们希望它们在完整的ENCODE项目中更加有用。我们还将提供全基因组多重比对的质量保证和汇总指标。这个领域有许多复杂的统计、算法和工程问题，我们将使用最先进的技术来解决这些问题。总的来说，我们的目标是在AWG的指导下，与ENCODE联盟的其他成员密切合作，提供ENCODE数据的深度集成。

DESCRIPTION (provided by applicant): The ENCODE Data Analysis Center (EDAC) proposal aims to provide a flexible analysis resource for the ENCODE project. The ENCODE project is a large multi center project which aims to define all the functional elements in the human genome. This will be achieved using many different experimental techniques coupled with numerous computational techniques. A critical part in delivering this set of functional elements is the integration of data from multiple sources. The ED AC proposal aims to provide this integration. As proscribed by the RFA for this proposal, the precise prioritization for the EDAC's work will be set by an external group, the Analysis Working Group (AWG). Based on previous experience, these analysis methods will require a variety of techniques. We expect to have to apply sophisticated statistical models to the integration of the data, in particular mitigating the problems of the extensive heterogeneity and correlation of variables on the human genome. We have statistical experts who can use the large size of the human genome, coupled with a limited number of sensible assumptions to produce statistical techniques which are robust to this considerable heterogeneity. We also expect to apply machine learning techniques to build integration methods combining datasets. These included Bayesian based inference methods and the robust computer science technique of Support Vector Machines. Each of these methods have performed well in the ENCODE pilot project and we expect them to be even more useful in the full ENCODE project. We will also provide quality assurance and summary metrics of genome-wide multiple alignments. This area has a number of complex statistical, algorithmic and engineering issues, which we will solve using state of the art techniques. Overall we aim to provide deep integration of the ENCODE data, under the direction of the AWG and in tight collaboration with the other members of the ENCODE consortium.