Develop a Dynamic Model that Incoporates Text-mining to Reconstruct Networks

开发结合文本挖掘来重建网络的动态模型

• 批准号: 7348388
• 负责人: CHRISTINA CHAN
• 金额: $ 25.05万
• 依托单位: MICHIGAN STATE UNIVERSITY
• 依托单位国家: 美国
• 财政年份: 2007
• 资助国家: 美国
• 起止时间: 2007-02-01 至 2011-01-31
• 项目状态: 已结题
• 来源: https://reporter.nih.gov/project-details/7348388
• 关键词: Algorithms; Animals; Cell Culture System; Cell physiology; Cells; Computer Simulation; Condition; Cultured Cells; Data; Diet; Disease; Drug Delivery Systems; Environment; Fatty acid glycerol esters; Feedback; Gene Expression; Gene Silencing; Genes; Inflammation; Knowledge; Liver; Measures; Metabolic; Modeling; Numbers; Ontology; Pathway interactions; Phenotype; RNA Interference; Rattus; Regulator Genes; Series; Structure; Text; Time; Transfection; Triglycerides; computer based statistical methods; cost effectiveness; desire; feeding; improved; in vivo; nonalcoholic steatohepatitis; novel strategies; reconstruction; research study; response; text searching

DESCRIPTION (provided by applicant): Improved understanding of disease mechanisms and drug target identification requires better understanding of how diseases alter cellular processes from the healthy state. Our group has developed an integrative pathway search algorithm that reconstructs networks of active pathways from gene expression and phenotypic profiles. Preliminary studies illustrate that this framework is able to reconstruct networks that include those pathways that should be altered, and how they should be altered, to obtain a desired phenotype. Nevertheless, the current framework may not fully capture transients, such as cycles and feedback loops. Furthermore, cells continuously reprogram gene regulatory networks as they sense changes in their environment. To understand how cells are regulated in response to environmental alterations, time series (i.e., dynamic) data are required. Correspondingly, a dynamic model is required to uncover the mechanisms from time series data. We hypothesize that incorporating domain knowledge and metabolic data into a dynamic model would enhance the accuracy of the genes chosen and in turn improve the prediction of the reconstructed networks. Unlike previous studies that have focused on using the gene ontology information, we propose to incorporate domain knowledge retrieved from the free text. This is significant because a large portion of the genes do not have gene ontological keywords. Additionally, it is often difficult to assess the accuracy of the network structures that have been inferred from experimental data because the underlying "true" regulatory network is unknown or unavailable a priori. Therefore, one needs to have a known network structure that can be used to optimize and evaluate the modeling frameworks. Once so optimized, the static and dynamic models will be applied to an experimental cell culture system, which has a perturbed (transfected or silenced) gene, and assessed as to how well each model predicts the resulting, measured phenotypic responses. The cost-effectiveness of the cell culture, in contrast to in vivo animal studies, allows us to establish, with experimental data, which model produces predictions of greater confidence. Having established which model is more predictive, we will apply that model to rats that are maintained on high fat diets, so as to identify the pathways that could be altered to reduce triglyceride storage (steatosis) and inflammation in the livers of these rats. The findings could have implications for identifying potential therapies for steatosis and, perhaps, even non-alcoholic steatohepatitis (NASH). The objectives will be achieved through the following aims: 1) Develop a novel approach that incorporates domain knowledge retrieved from the free text as well as gene expression data to predict cellular or phenotypic responses. 2) Develop an optimized dynamic Bayesian Network to infer gene regulatory networks from time series data. 3) Experimentally validate the model predictions for the cell culture system. 4) Characterize the livers from rats fed high fat vs. normal diets.

描述（由申请人提供）：为了更好地理解疾病机制和药物靶点识别，需要更好地理解疾病如何改变健康状态下的细胞过程。我们的小组开发了一种综合途径搜索算法，可以根据基因表达和表型谱重建活性途径网络。初步研究表明，该框架能够重建网络，其中包括那些应该改变的途径，以及如何改变它们，以获得所需的表型。然而，当前的框架可能无法完全捕获瞬态，例如周期和反馈循环。此外，细胞在感知环境变化时不断重新编程基因调控网络。为了了解细胞如何响应环境变化而受到调节，需要时间序列（即动态）数据。相应地，需要一个动态模型来揭示时间序列数据的机制。我们假设将领域知识和代谢数据纳入动态模型将提高所选基因的准确性，进而改善重建网络的预测。与之前专注于使用基因本体信息的研究不同，我们建议整合从自由文本中检索的领域知识。这很重要，因为很大一部分基因没有基因本体论关键词。此外，通常很难评估从实验数据推断出的网络结构的准确性，因为潜在的“真实”监管网络是未知的或先验不可用的。因此，需要有一种已知的网络结构，可用于优化和评估建模框架。一旦如此优化，静态和动态模型将应用于实验细胞培养系统，该系统具有扰动（转染或沉默）基因，并评估每个模型预测所得的、测量的表型反应的程度。与体内动物研究相比，细胞培养的成本效益使我们能够利用实验数据建立哪种模型可以产生更有信心的预测。在建立了哪种模型更具预测性后，我们将该模型应用于维持高脂肪饮食的大鼠，以确定可以改变的途径以减少这些大鼠肝脏中的甘油三酯储存（脂肪变性）和炎症。这些发现可能对确定脂肪变性甚至非酒精性脂肪性肝炎（NASH）的潜在疗法具有重要意义。这些目标将通过以下目标实现：1）开发一种新颖的方法，结合从自由文本中检索到的领域知识以及基因表达数据来预测细胞或表型反应。 2）开发优化的动态贝叶斯网络，从时间序列数据推断基因调控网络。 3）通过实验验证细胞培养系统的模型预测。 4) 表征高脂肪大鼠与正常饮食大鼠的肝脏。