Aghdam Rosa, Ganjali Mojtaba, Zhang Xiujun, Eslahchi Changiz
Faculty of Mathematical Sciences, Department of Statistics, Shahid Beheshti University, G.C., Tehran, Iran.
Mol Biosyst. 2015 Mar;11(3):942-9. doi: 10.1039/c4mb00413b. Epub 2015 Jan 21.
Inferring Gene Regulatory Networks (GRNs) from gene expression data is a major challenge in systems biology. The Path Consistency (PC) algorithm is one of the popular methods in this field. However, as an order dependent algorithm, PC algorithm is not robust because it achieves different network topologies if gene orders are permuted. In addition, the performance of this algorithm depends on the threshold value used for independence tests. Consequently, selecting suitable sequential ordering of nodes and an appropriate threshold value for the inputs of PC algorithm are challenges to infer a good GRN. In this work, we propose a heuristic algorithm, namely SORDER, to find a suitable sequential ordering of nodes. Based on the SORDER algorithm and a suitable interval threshold for Conditional Mutual Information (CMI) tests, a network inference method, namely the Consensus Network (CN), has been developed. In the proposed method, for each edge of the complete graph, a weighted value is defined. This value is considered as the reliability value of dependency between two nodes. The final inferred network, obtained using the CN algorithm, contains edges with a reliability value of dependency of more than a defined threshold. The effectiveness of this method is benchmarked through several networks from the DREAM challenge and the widely used SOS DNA repair network in Escherichia coli. The results indicate that the CN algorithm is suitable for learning GRNs and it considerably improves the precision of network inference. The source of data sets and codes are available at .
从基因表达数据推断基因调控网络(GRN)是系统生物学中的一项重大挑战。路径一致性(PC)算法是该领域常用的方法之一。然而,作为一种依赖顺序的算法,PC算法并不稳健,因为如果基因顺序被打乱,它会得到不同的网络拓扑结构。此外,该算法的性能取决于用于独立性检验的阈值。因此,为PC算法的输入选择合适的节点顺序和适当的阈值是推断良好GRN的挑战。在这项工作中,我们提出了一种启发式算法,即SORDER,以找到合适的节点顺序。基于SORDER算法和用于条件互信息(CMI)检验的合适区间阈值,开发了一种网络推断方法,即共识网络(CN)。在所提出的方法中,对于完全图的每条边,定义一个加权值。该值被视为两个节点之间依赖关系的可靠性值。使用CN算法获得的最终推断网络包含依赖关系可靠性值超过定义阈值的边。通过来自DREAM挑战的几个网络以及大肠杆菌中广泛使用的SOS DNA修复网络对该方法的有效性进行了基准测试。结果表明,CN算法适用于学习GRN,并且显著提高了网络推断的精度。数据集和代码的来源可在 获得。