Breaking the Dimensionality Barrier
Breaking the Dimensionality Barrier
复制标题
DOI：
10.1007/978-1-61737-950-5_2
复制
发表时间：
2011-01-01
期刊：
影响因子：
--
通讯作者：
Bagwell, C. Bruce
中科院分区：
文献类型：
--
作者：
Bagwell, C. Bruce
生物技术的最新进展导致细胞仪能够进行大量相关的细胞测量，通常超过十个。在不久的将来，这个数字可能会增加五倍甚至更高。基于检查一个测量值与另一种测量值的传统分析策略不适合高维数据分析，因为测量组合的数量随维度而形成了一种复杂性障碍。这种维度障碍限制了细胞术和其他技术在高维数据中嵌入的重要信息可视化和分析其最大潜力概率状态建模（PSM）。这个新系统基于与所有测量相关的概率创建虚拟进程变量。 PSM可以产生一个图表，该图传达了比数百个传统直方图传达更多有关样品的信息。这些PSM覆盖层揭示了细胞分化时表型变化的丰富相互作用。最终结果是对复杂种群中本体论过程的分子遗传基础（例如在骨髓和外周血中发现）的分子遗传基础的更深入。高维数据的分析和可视化。
Recent advances in biotechnology have resulted in cytometers capable of performing numerous correlated measurements of cells, often exceeding ten. In the near future, it is likely that this number will increase by fivefold and perhaps even higher. Traditional analysis strategies based on examining one measurement versus another are not suitable for high-dimensional data analysis because the number of measurement combinations expands geometrically with dimension, forming a kind of complexity barrier. This dimensionality barrier limits cytometry and other technologies from reaching their maximum potential in visualizing and analyzing important information embedded in high-dimensional data.This chapter describes efforts to break through this barrier and allow the visualization and analysis fatly number of measurements with a new paradigm called Probability State Modeling (PSM). This new system creates a virtual progression variable based on probability that relates all measurements. PSM can produce a single graph that conveys more information about a sample than hundreds of traditional histograms. These PSM overlays reveal the rich interplay of phenotypic changes in cells as they differentiate. The end result is a deeper appreciation of the molecular genetic underpinnings of ontological processes in complex populations such as found in bone marrow and peripheral blood.Eventually these models will help investigators better understand normal and abnormal cellular progressions and will be a valuable general tool for the analysis and visualization of high-dimensional data.