Development of integrated web interfaces for Bioconductor genomic data analysis annotation and visualization tools

基因组数据，特别是微阵列表达谱研究，以巨大的数字矩阵的形式出现，从10,000到6,000,000行到数百到数千列。需要对这些数据进行转换、标准化、可视化和注释。这些矩阵的行报告了基因在不同条件下的活性（表达水平）。庞大的数据量，以及描述这些实验数据所涉及的复杂性，导致了基于阵列的高通量基因组数据的几个主要公共存储库的创建：GEO （NCBI，美国）和ArrayExpress （EBI，剑桥，英国）。我们在EBI的团队还开发了Expression Profiler (EP)，这是一个基于web的探索性数据分析平台，可以提供一些关于ArrayExpress公共数据的基本见解。科学界在创建处理如此大规模数据的工具方面的主要工作集中在一组统称为Bioconductor的开源命令行驱动工具中。这些工具或“包”是由专业应用领域的领导者开发的：标准化（使来自不同实验室的数据具有可比性的数学方法），信号通路分析，聚类分析，元分析等，因此是前沿功能基因组学分析技术的事实上的标准。与此同时，总的来说，Bioconductor的唯一用户仍然是老练的生物信息学家，而湿实验室生物学家（产生实际数据的实验家）发现R环境的学习曲线太陡峭而无法学习，R语言太复杂而无法掌握，命令行灵活性细节太令人生畏。此外，即使在Bioconductor内部，不同的封装也提供不同的、通常不兼容的处理数据输入、输出和交换的范例。对于参与功能基因组学和蛋白质组学实验研究的生物学家来说，有一个明确的、明确的需求，就是提供方便的途径来使用Bioconductor的力量。该项目建议利用EP分析框架开发一套标准的、统一的基于web的界面，以连接核心Bioconductor模块，该模块也将利用ArrayExpress数据库。拟议的系统将使生物学家能够安全地上传他们的实验数据，用最好的生物导体算法分析它们，并将它们与存储库中相关的公共高通量数据进行比较或分析。数据分析程序将利用EBI可用的高性能计算基础设施，结果将存储在系统中，可以通过网络浏览器从世界任何地方访问。另一个独特的优势是将Bioconductor软件包集成在一组web接口中：这些接口也可以作为web服务访问，即可以合并到自动数据分析工作流中。换句话说，即使是老练的生物信息学家也可能会发现这个系统很有用（见所附的支持信）。

Genomic data, particularly microarray expression profiling studies, comes in the shape of huge matrices of numbers, anywhere from 10,000 to 6,000,000 rows by hundreds to thousands of columns. These data need to be transformed, standardized, visualized, and annotated. The rows of these matrices report activity (expression levels) of genes under various conditions. The huge data volume, as well as the complexity involved with describing such experimental data, resulted in the creation of a few major public repositories for array-based high throughput genomics data: GEO (NCBI, USA) and ArrayExpress (EBI, Cambridge, UK). Our group at the EBI also has developed Expression Profiler (EP), a web-based platform for exploratory data analysis, which can provide some basic insights into the public data in ArrayExpress. The major thrust of the scientific community's work in creating tools for dealing with such large-scale data has concentrated within the set of open source command-line driven tools collectively called Bioconductor. These tools, or 'packages', are developed by leaders in specialized areas of application: normalization (mathematical methods of making data coming from different laboratories comparable), signalling pathway analysis, clustering analysis, meta-analysis, etc., and are therefore the de facto standard for cutting-edge functional genomics analysis technologies. At the same time, by and large the only users of Bioconductor remain the sophisticated bioinformaticians, while wet-lab biologists (experimentalists who produce the actual data) find the learning curve of the R environment too steep to learn, the R language too complex to master, and the command-line flexibility details too daunting. Moreover, even within Bioconductor, different packages offer different, often incompatible, paradigms of dealing with the data input, output and interchange. There is a definite, clear need to provide easy access to the power of Bioconductor for biologists involved in functional genomics and proteomics experimental research. This project proposes to utilise the EP analytical framework to develop a set of standard, unified look-and-feel web-based interfaces to core Bioconductor modules, which will also make use of the ArrayExpress database. The proposed system will enable biologists to upload securely their experimental data, analyse them with the best available Bioconductor algorithms and to compare or analyse them together with related public high-throughput data in the repository. The data analysis routines will take advantage of the high-power computing infrastructure available at the EBI, and the results will be stored within the system, accessible form anywhere in the world via a web-browser. A further unique advantage is provided by the integration of Bioconductor packages within a set of web interfaces: the interfaces can also be accessed as Web Services, i.e. can be incorporated in automatic data analysis workflows. In other words, even sophisticated bioinformaticians are likely to find this system useful (see attached letters of support).