CIF: Small: Sparse Signal Processing Methods for Inference of Differential Gene Regulatory Networks

活细胞中的基因相互作用，形成一个复杂的网络，调节分子功能和生物过程。最近的研究表明，当细胞对不同的环境信号做出反应时，基因调控网络(GRN)可以经历实质性的重新布线。更广泛地说，GRN的不同变化可能取决于环境、组织类型、疾病状态、发育和物种。因此，识别GRN中的这种变化是充分理解GRN如何工作的重要一步，并可能对理解各种生物过程和生物医学研究产生深远影响。然而，目前的计算方法主要集中在单一条件下GRN的推断或试图识别差异基因的共表达。目前还没有一种方法能够系统地确定GRN的结构变化。这个项目的目标是设计系统的方法，可以利用基因表达和其他相关数据来推断GRN的结构变化，并开发实现这种推断方法的软件。智力优势：建议的推理方法寻求利用GRN的一个重要属性，即GRN的稀疏性及其结构变化，它们建立在稀疏信号处理的创新思想之上。他们使用新的推理技术将各种类型的数据集成到一个统一的框架中。这种系统的方法有望显著提高推理精度，而不是现有的特别方法。所提出的方法和由此产生的软件将为差异GRN的分析提供有价值的工具，并有助于揭示不同条件下的差异基因-基因相互作用。虽然对于微分GRN的推理问题已经建立了数学公式，但其背后的数学思想和相关的推理方法也可能在更广泛的稀疏信号处理和稀疏模型学习领域得到应用。广泛的影响：该项目的成功完成将对信号处理、系统生物学和医学领域产生广泛的影响。这将为稀疏信号处理领域提供一套新颖的推理方法。这也将有助于理解GRN在各种疾病状态下、细胞分化过程中或细胞对环境变化的反应中的动态变化。特别是，了解癌症等遗传性疾病中不同的基因-基因相互作用将为新的诊断和治疗方法的开发打开大门。总体而言，拟议的研究将有助于发现新知识，并可能在医学研究中找到应用，从而造福整个社会。拟议的项目还将对研究生和本科生的跨学科教育产生积极影响，并吸引少数族裔学生参与尖端研究。

Genes in living cells interact with each other and form a complex network to regulate molecular functions and biological processes. Recent studies have shown that gene regulatory networks (GRNs) can undergo substantial rewiring when cells respond to different environmental signals. More generally, differential changes of GRNs can occur depending on environment, tissue type, disease state, development and speciation. Identification of such changes in GRNs is thus an important step to fully understand how GRNs work, and may have profound impact on the understanding of various biological processes and on biomedical research. However, current computational methods are focused on the inference of GRNs under a single condition or attempt to identify differential gene co-expression. No method currently exists that is capable of systematically identify structural changes in GRNs. The objectives of this project are to design systematic methods that can exploit gene expression and other relevant data to infer structural changes of GRNs, and to develop software that implements such inference methods. Intellectual merit: The proposed inference methods seek to leverage an important attribute of GRNs, the sparsity in both GRNs and their structural changes, and they are built on the innovative ideas of sparse signal processing. They employ novel inference techniques to integrate various types of data into a unified framework. This systematic approach is expected to significantly improve inference accuracy over existing ad hoc methods. The proposed methods and the resulting software will provide a valuable tool for the analysis of differential GRNs and help to uncover differential gene-gene interactions under different conditions. Although mathematical formulations are framed for the problem of inference of differential GRNs, the underlying mathematical ideas and the associated inference methods may also find applications in more general domain of sparse signal processing and sparse model learning.Broader impact: Successful completion of the proposed project will have a broad impact on the fields of signal processing, systems biology, and medicine. It will contribute a novel set of inference methods to the field of sparse signal processing. It will also help to understand the dynamic changes of GRNs that occur in various disease states, during cell differentiation, or in cellular response to environmental changes. In particular, understanding differential gene-gene interactions in genetic diseases such as cancer will open the door to new diagnostics and therapeutics development. Overall, the proposed research will help to discover new knowledge and potentially find applications in medical research, thereby benefiting society as a whole. The proposed project will also positively impact interdisciplinary education at both graduate and undergraduate levels, and attract minority students to be involved in cutting edge research.