Flexible and Robust Nonlinear Statistical Modeling Based on High-Dimensional Complex Heterogeneous Data Analysis

我们开发了强大的统计方法，从各种类型的高维异构数据中提取有价值的信息。作为结果，当一个子集的特征变量被定义的先验，我们开发了一系列的统计方法，可以评估是否有一个独特的分布，通过比较在给定的大数据集的背景变量的子集。我们还提出了一种基于微阵列基因表达数据和转录结合位点等其他类型数据的大基因网络鲁棒估计方法，并将其应用于癌症异质性分析。

We developed robust statistical methods for extracting valuable information from various types of high-dimensional heterogeneous data. As the results, when a subset of feature variables is defined a priori, we develop a series of statistical methods that can evaluate whether the subset has a unique distribution by comparing with background variables in the given large dataset. We also proposed a robust estimation method for large gene network from microarray gene expression data and other types of data like transcription binding sites and applied it to the analysis of cancer heterogeneity.