• 主题
高级搜索  >
历史搜索 清空历史
首页> 外文期刊>Computational Biology and Bioinformatics, IEEE/ACM Transactions on >Continuous Cotemporal Probabilistic Modeling of Systems Biology Networks from Sparse Data
【24h】

Continuous Cotemporal Probabilistic Modeling of Systems Biology Networks from Sparse Data

机译：基于稀疏数据的系统生物学网络的连续同期概率建模

获取原文
获取原文并翻译 | 示例
           
页面导航
  • 摘要
  • 著录项
  • 相似文献
  • 相关主题

摘要

Modeling of biological networks is a difficult endeavor, but exploration of this problem is essential for understanding the systems behavior of biological processes. In this contribution, developed for sparse data, we present a new continuous Bayesian graphical learning algorithm to cotemporally model proteins in signaling networks and genes in transcriptional regulatory networks. In this continuous Bayesian algorithm, the correlation matrix is singular because the number of time points is less than the number of biological entities (genes or proteins). A suitable restriction on the degree of the graph's vertices is applied and a Metropolis-Hastings algorithm is guided by a BIC-based posterior probability score. Ten independent and diverse runs of the algorithm are conducted, so that the probability space is properly well-explored. Diagnostics to test the applicability of the algorithm to the specific data sets are developed; this is a major benefit of the methodology. This novel algorithm is applied to two time course experimental data sets: 1) protein modification data identifying a potential signaling network in chondrocytes, and 2) gene expression data identifying the transcriptional regulatory network underlying dendritic cell maturation. This method gives high estimated posterior probabilities to many of the proteins' directed edges that are predicted by the literature; for the gene study, the method gives high posterior probabilities to many of the literature-predicted sibling edges. In simulations, the method gives substantially higher estimated posterior probabilities for true edges and true subnetworks than for their false counterparts.
机译：对生物网络进行建模是一项艰巨的努力，但是对这个问题的探索对于理解生物过程的系统行为至关重要。在为稀疏数据开发的这项贡献中，我们提出了一种新的连续贝叶斯图形学习算法，可以同时模拟信号网络中的蛋白质和转录调控网络中的基因。在这种连续贝叶斯算法中，相关性矩阵是奇异的，因为时间点的数量少于生物实体（基因或蛋白质）的数量。应用对图的顶点程度的适当限制，并以基于BIC的后验概率得分为指导的Metropolis-Hastings算法。进行了十次独立且多样的算法运行，因此可以很好地探索概率空间。开发了用于测试算法对特定数据集的适用性的诊断程序；这是该方法的主要优点。这种新颖的算法应用于两个时程实验数据集：1）识别软骨细胞中潜在信号网络的蛋白质修饰数据，以及2）识别树突状细胞成熟的转录调控网络的基因表达数据。这种方法为文献所预测的许多蛋白质的定向边缘提供了很高的估计后验概率。对于基因研究，该方法为许多文献预测的同胞边缘提供了较高的后验概率。在仿真中，该方法给出的真实边缘和真实子网络的估计后验概率要比其假对等概率高得多。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱：kefu@zhangqiaokeyan.com

京公网安备：11010802029741号 ICP备案号：京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有

  • 客服微信

  • 服务号