Flexible statistical machine learning techniques for cancer-related data

描述(由申请人提供)：基因表达提供了促进肿瘤恶性的细胞变化的快照。定量的基因表达分析，特别是通过DNA微阵列实现的，已经发现了许多新的重要的癌症相关基因，并导致了新的基于基因组的临床测试的发展。在基因表达分析的定量方面，许多统计学方法被用来研究人类肿瘤，并将它们分类，以用于预测临床行为。尽管取得了进展，但随着技术的快速进步，癌症研究中产生了大量复杂的数据。分析这些数据变得越来越具有挑战性。这些挑战呼唤新的统计学习方法，特别是对于高维和噪声数据。这个项目的目标是开发一系列新的统计学习技术来解决复杂的学习问题。特别是，该项目开发了(1)评估高维数据聚类的统计意义的新技术；(2)几种新的预测模型，包括分类和回归，有望产生高度竞争的准确性和可解释性；(3)高维生物标记物/变量选择的新方法；(4)估计用于生物网络构建的高维协方差/精度矩阵的新方法。预计这些新的发展将使科学家能够以准确的预测精度和更高的可解释性来分析复杂的癌症基因组数据。研究小组将把拟议的技术应用于癌症研究数据分析。该项目的成功将在统计机器学习和癌症研究之间架起一座桥梁。 公共卫生相关性：该项目旨在开发一系列新的统计学习技术，以解决复杂的学习问题，特别是具有高维和噪声数据的问题，如基因表达数据。预计这些新技术将使科学家能够以准确的预测精度和更高的可解释性分析复杂的癌症基因组数据。

DESCRIPTION (provided by applicant): Gene expression provides a snapshot of the cellular changes that promote tumor malignancy. Quantitative gene expression analysis, especially as implemented by DNA microarrays, has identified many new important cancer related genes and led to the development of new genomic-based clinical tests. For the quantitative aspect of gene expression analysis, many statistical methods have been used to study human tumors and to classify them into groups that can be used to predict clinical behavior. Despite progress, with the rapid advance of technology, massive and complex data are being generated in cancer research. Analyzing such data becomes more and more challenging. These challenges call for novel statistical learning methods, especially for high dimensional and noisy data. The goal of this project is to develop a host of new statistical learning techniques for solving complicated learning problems. In particular, this project develops (1) novel techniques to assess statistical significance of clustering for high dimensional data; (2) several novel predictive models including classification and regression which are expected to yield highly competitive accuracy and interpretability; (3) new methods for high dimensional biomarker/variable selection; (4) new approaches to estimate high dimensional covariance/precision matrix for biological network construction. These new developments are expected to allow scientists to analyze complex cancer genomic data with accurate prediction accuracy and increased interpretability. The research team will apply the proposed techniques to cancer research data analysis. The success of this project will be important in bridging statistical machine learning and cancer research. PUBLIC HEALTH RELEVANCE: This project aims to develop a host of new statistical learning techniques for solving complicated learning problems, especially for problems with high dimensional and noisy data such as gene expression data. These new techniques are expected to allow scientists to analyze complex cancer genomic data with accurate prediction accuracy and increased interpretability.