Equivalent Partial Correlation Methods for Integrative Genetic Network Analysis

 描述（由申请人提供）：高通量技术的出现使得同时测量数千个基因的活性成为可能，这为科学家推断全球基因调控网络（GRNs）提供了重要机会。GRNs的准确推断对于系统地了解其分子机制、阐明细胞加工过程失调导致疾病的发生机制以及确定潜在的疾病治疗靶点具有重要意义。由于高吞吐量数据的高维性和高复杂性，全局GRNs的推理在很大程度上依赖于计算方法的进步。 然而，目前的计算方法推断的全球GRNs要么是不准确的或计算上不可行的。如何推断全球GRN对现有的统计方法提出了很大的挑战。 研究人员提出了一个等效的偏相关系数的测量，并在此基础上开发了一个创新的计算框架，推断全球GRNs。这种新的偏相关系数测度可以用一个简化的条件集来计算，因此对于高维问题是可行的。在新的框架下，研究人员开发了一系列算法，这些算法能够通过整合各种类型的高通量分子分析数据，例如，mRNA表达、拷贝数变异、甲基化、microRNA和蛋白质表达，并用各种临床协变量进行调整，例如，年龄、性别和疾病阶段。所提出的算法被应用于推断各种类型的癌症的全局GRNs，特别关注肺癌，而它们适用于所有其他癌症。 疾病的种类。在统计上，该项目提出了一个创新的框架来推断全局GRNs，在此框架下开发的算法不仅计算效率高，而且在数据集成，协变量调整，先验知识集成和网络比较方面非常灵活。在生物医学上，这是第一项使用严格的统计方法整合此类全面和互补信息的研究，以研究各种类型癌症的全球GRNs。

 DESCRIPTION (provided by applicant): The emergence of high-throughput technologies has made it feasible to measure the activities of thousands of genes simultaneously, which provides scientists a major opportunity to infer global gene regulatory networks (GRNs). Accurate inference of GRNs is very important, which allows people to gain a systematic understanding of the molecular mechanism, to shed light on the mechanism of diseases that occur when cellular processed are dysregulated, and furthermore to identify potential therapeutic targets for diseases. Given the high dimensionality and high complexity of high-throughput data, inference of global GRNs largely relies on the advance of computational methods. However, the current computational methods for inference of global GRNs are either inaccurate or computationally infeasible. How to infer global GRNs has put a great challenge on the current statistical methodology. The investigators propose an equivalent measure of partial correlation coefficients, and based on which develop an innovative computational framework for inference of global GRNs. The new measure of partial correlation coefficients can be evaluated with a reduced conditional set and thus feasible for high dimensional problems. Under the new framework, the investigators develop a series of algorithms which are able to provide a comprehensive inference for global GRNs by integrating various types of high-throughput molecular profiling data, e.g., mRNA expression, copy number variation, methylation, microRNA, and protein expression, and adjusting with various clinical covariates, e.g., age, gender, and disease stage. The proposed algorithms are applied to infer the global GRNs for various types of cancer with a special focus on lung cancer, while they are applicable to all other types of diseases. Statistically, this project proposes an innovative framework for inference of global GRNs, and the algorithms developed under which are not only computationally efficient, but also very flexible in data integration, covariate adjustment, prior knowledge integration, and network comparison. Biomedically, this is the first study to integrate such comprehensive and complementary information using rigorous statistical methods to study the global GRNs for various types of cancer.