Isomap在基因表达谱数据聚类分析中的应用

基因表达谱数据的聚类分析对于研究基因功能和基因调控机制有重要意义。基于非线性降维算法等容特征映射 ,提出了一种新的大规模基因表达谱数据聚类算法 ,该方法改进了样本向量之间的距离度量 ,用测地距离代替传统的欧式距离 ,有助于挖掘高维数据内在的几何结构。将该算法应用于两个公开的基因表达数据集 ,并用一种新的评价方法Normalized Cut将聚类结果与其他聚类方法的结果进行了比较。结果表明 ,该文的聚类算法优于其他聚类算法 ,聚类结果具有明显的生物学意义 ,并能对数据的类别数作出较好的预测和评估

Application of isomap for cluster analyses of gene expression data

The cluster analysis of gene expression data is an important means for discovering gene functions and regulatory mechanisms. A novel clustering algorithm based on isometric feature mapping (Isomap) was used to analyze large-scale gene expression data. The algorithm improved the distance measurement between samples by replacing the classical Euclidean distance with the geodesic distance, which helped preserve the intrinsic geometry of the high-dimensional data. Two public gene expression datasets were analyzed, with the results evaluated using a new method named Normalized Cut. The algorithm provides more useful information than other methods, with obvious biological meanings. The optimal cluster number can also be estimated by the method.