Identifying novel molecular insights from life science massive data

生物数据集的规模和复杂性呈指数级增长，使它们成为大数据世界的一部分。数以百计的知识库组织生物学研究，并向生物学家提供结果，以便进行额外的分析。目前，由于不兼容的数据类型以及将数据分离到无法组合的专用存储库中，这些存储库中的数据无法得到最佳利用。因此，需要在许多数据库中进行导航才能进行搜索，而研究人员并不总是具有整合不同数据源的专业知识或计算知识。研究问题通过高通量技术（即测序和微阵列）获得的基因表达实验被广泛用于获得基因签名以定义特定表型。为了有价值，基因签名需要提供有关特定分子过程和更大的系统视图上下文的信息。此外，这些签名需要超越数据集或数据库的简单集成，以提供对新数据的生物学功能和途径的有效和准确的见解。为了满足这些需求，我们开发了Kibio。这是一个结合了Elasticsearch（一个搜索引擎）和Kibi（一个网络可视化插件）的新门户网站，用于创建生物数据库之间的关系搜索。通过该提案，我们将实现分析工具，以利用高通量数据，临床和病理信息。impactskibio.science是一个集中式平台，可以轻松地可视化，解释和执行多种数据类型的生物分析。本研究计划的主要贡献将是提供分析工具，以促进功能解释和实验数据的可视化（例如基因签名）。我们将创建新的统计模型来测量基因标记的质量和准确性，并提供生物分子共提到/相关性的统计数据。进一步提高kibio的分析能力。科学，专门用于临床或研究目的的动态可视化将被实施。结合新的分析工具，这将提供将大量生命科学数据转化为可操作知识的方法。

BackgroundBiological datasets have grown exponentially in size and complexity making them part of the world of big data. Hundreds of repositories organize biological research and make results available to biologists to be subjected to additional analyses. At the moment, the data from these repositories cannot be optimally exploited because of incompatible data types and of the separation of data into specialized repositories that cannot be combined. Therefore, navigation through many databases is required to perform searches and researchers do not always have the expertise or computational knowledge to integrate the different sources of data.Research ProblemGene expression experiments obtained from high throughput technologies (i.e. sequencing and microarray) are widely used to obtain gene signatures to define a particular phenotype. To be valuable, gene signatures are required to provide information on specific molecular process involved and a larger system-view context. In addition, these signatures need to go beyond the simple integration of datasets or databases, to provide valid and accurate insight on biological function and pathways of new data. To answer these needs, we have developed Kibio.science, a new portal that combines Elasticsearch, a search engine, and Kibi, a web visualization plugin, to create the relational searches between biological databases. With this proposal, we will implement analytical tools to take advantage of high-throughput data, clinical and pathological information.ImpactsKibio.science is a centralized platform that make it easy to visualize, interpret and perform biological analysis of multiple data types. The major contribution of this research program will be to provide analytical tools that facilitate functional interpretation and visualization of experimental data (e.g. gene signatures). We will create new statistical models to measure quality and accuracy of gene signatures in addition to provide statistics on co-mentions/correlations of biomolecules. Furthermore, to enhance the analytical power of kibio.science, specialized dynamic visualizations dedicated to clinical or research purposes will be implemented. Combined with new analytic tools, this will provide ways of translating massive life science data into actionable knowledge.