VISUAL AND ANALYTICAL TOOLS FOR CLUSTER ANALYSIS

这个子项目是许多研究子项目中的一个 由NIH/NCRR资助的中心赠款提供的资源。子项目及 研究者（PI）可能从另一个NIH来源获得了主要资金， 因此可以在其他CRISP条目中表示。所列机构为 研究中心，而研究中心不一定是研究者的研究机构。 聚类被认为是发现记录（如基因、网络或其他生物医学数据）之间关系的最强大的技术之一。由于存在大量的聚类算法，并且无法在计算上选择适合各种情况和数据集的单个算法，因此需要确定解决这个问题的方法。数据的指数增长加剧了这个问题，这推动了对解决问题的新方法的需求。我们的工作解决了不同算法和组中记录之间的关系问题，并使科学家能够查看多个结果，以得出有意义的结论。我们创建了新的分析和可视化工具和技术，以深入了解单个和多个聚类算法的结果。我们的工作有利于利用可视化技术进行数据表示的高维数据的分析。我们的新的聚类措施提供了信息的接近度，并提供了一种手段，包括最大数量的记录之间的共同集群成员的记录的视觉投影。 我们继续开发我们的可视化和分析方法的工具套件。我们在系统中集成了新的可视化，以及算法和其他交互方法的实现。我们为套件间通信实现了一种传播机制，并测试了各种数据集，以确定这些工具的可伸缩性。我们将我们的工具与其他算法进行了比较，其中没有一种算法能够使用单一的技术（视觉和/或分析）来呈现记录和组比较。我们利用我们的工具在几个研究从初步到验证性的研究，我们展示了膀胱移行细胞癌的数据集的案例研究，结合统计分析，可视化和数据挖掘。

This subproject is one of many research subprojects utilizing the resources provided by a Center grant funded by NIH/NCRR. The subproject and investigator (PI) may have received primary funding from another NIH source, and thus could be represented in other CRISP entries. The institution listed is for the Center, which is not necessarily the institution for the investigator. Clustering is considered one of the most powerful techniques utilized in discovery of relationships among records such as genes, networks, or other biomedical data. Due to the large number of clustering algorithms in existence and inability to computationally choose a single algorithm that would suit a variety of situations and data sets, there is a need to identify ways to tackle this problem. This problem is exacerbated by the exponential growth of data, which drive the need for new approaches to the problem. Our work addresses questions of relationships among records in different algorithms and groups, and enables scientists to look through multiple results in order to draw meaningful conclusions. We created new analytical and visual tools and techniques to provide insights into single and multiple clustering algorithm results. Our work facilitates the analysis of high-dimensional data utilizing visual techniques for data presentation. Our new clustering measure gives information on proximity and provides a means for visual projection of the records including the largest number of common cluster memberships among the records. We continue with the development of our suite of tools with visual and analytical approaches. We integrated new visualizations into the system, together with implementation of algorithms and additional interaction methods. We implemented a propagation mechanism for the inter-suite communication and tested a variety of data sets to determine scalability of these tools. We compared our tools with other algorithms, none of which are capable of presenting record and group comparisons using a single technique, visually and/or analytically. We utilize our tools in several research studies from preliminary to confirmatory and we show a case study of a transitional cell carcinoma of the bladder data set, combining statistical analysis, visualization and data mining.