Moffa Giusi, Erdmann Gerrit, Voloshanenko Oksana, Hundsrucker Christian, Sadeh Mohammad J, Boutros Michael, Spang Rainer
Department of Statistical Bioinformatics, Institute of Functional Genomics, University of Regensburg, Regensburg, Germany.
Division of Signaling and Functional Genomics, German Cancer Research Center (DKFZ) and Department of Cell and Molecular Biology, Faculty of Medicine Mannheim, Heidelberg University, Heidelberg, Germany.
PLoS One. 2016 Jun 1;11(6):e0155999. doi: 10.1371/journal.pone.0155999. eCollection 2016.
Cellular signalling pathways consolidate multiple molecular interactions into working models of signal propagation, amplification, and modulation. They are described and visualized as networks. Adjusting network topologies to experimental data is a key goal of systems biology. While network reconstruction algorithms like nested effects models are well established tools of computational biology, their data requirements can be prohibitive for their practical use. In this paper we suggest focussing on well defined aspects of a pathway and develop the computational tools to do so. We adapt the framework of nested effect models to focus on a specific aspect of activated Wnt signalling in HCT116 colon cancer cells: Does the activation of Wnt target genes depend on the secretion of Wnt ligands or do mutations in the signalling molecule β-catenin make this activation independent from them? We framed this question into two competing classes of models: Models that depend on Wnt ligands secretion versus those that do not. The model classes translate into restrictions of the pathways in the network topology. Wnt dependent models are more flexible than Wnt independent models. Bayes factors are the standard Bayesian tool to compare different models fairly on the data evidence. In our analysis, the Bayes factors depend on the number of potential Wnt signalling target genes included in the models. Stability analysis with respect to this number showed that the data strongly favours Wnt ligands dependent models for all realistic numbers of target genes.
细胞信号通路将多种分子相互作用整合为信号传播、放大和调节的工作模型。它们被描述为网络并可视化。根据实验数据调整网络拓扑结构是系统生物学的一个关键目标。虽然像嵌套效应模型这样的网络重建算法是计算生物学中成熟的工具，但其数据要求在实际应用中可能过高。在本文中，我们建议专注于信号通路的明确方面，并开发相应的计算工具。我们调整了嵌套效应模型的框架，以关注HCT116结肠癌细胞中Wnt信号激活的一个特定方面：Wnt靶基因的激活是否依赖于Wnt配体的分泌，或者信号分子β-连环蛋白的突变是否使这种激活与其无关？我们将这个问题构建为两类相互竞争的模型：依赖Wnt配体分泌的模型与不依赖的模型。模型类别转化为网络拓扑中信号通路的限制。依赖Wnt的模型比不依赖Wnt的模型更灵活。贝叶斯因子是在数据证据基础上公平比较不同模型的标准贝叶斯工具。在我们的分析中，贝叶斯因子取决于模型中包含的潜在Wnt信号靶基因的数量。关于这个数量的稳定性分析表明，对于所有实际数量的靶基因，数据强烈支持依赖Wnt配体的模型。
