【摘要】 基因表达谱的可视化本质上是高维数据的降维问题。采用流形学习算法来解决基因表达谱的降维数据可视化,讨论了典型的流形学习算法(Isomap和LLE)在表达谱降维中的适用性。通过类内/类间距离定量评价数据降维的效果,对两个典型基因芯片数据集(结肠癌基因表达谱数据集和急性白血病基因表达谱数据集)进行降维分析,发现两个数据集的本征维数都低于3,因而可以用流形学习方法在低维投影空间中进行可视化。与传统的降维方法(如PCA和MDS)的投影结果作比较,显示Isomap流形学习方法有更好的可视化效果。更多还原

【Abstract】 The major problem of the visualization of gene expression profile is dimension reduction.In this paper,we discuss the adaptability of two classic methods(Isomap and LLE) of Manifold Learning in dimension reduction and perform an analysis of two typical microarray data(colon cancer and leukaemia gene expression datasets),in which the assessing the dimension reduction effect is quantitative by using the function of the within and the without class distance.The dimensionality of the two datasets was found less than three,so they can be visualized in low dimension space.Compared with the two traditional linear methods(PCA and MDS),the result shows that Isomap performs better in the gene expression profile visualization.更多还原