基于结构的方法可以识别生物启发的布尔网络集合中的驱动节点
Structure-based approach can identify driver nodes in ensembles of biologically-inspired Boolean networks
AI总结：
本研究扩展了生物启发布尔网络集合的分析，使用PRINCE传播、修改后的PRINCE传播和CheiRank三种结构指标对FVS子集进行排序，并验证了其准确性，发现传播指标结合FVS可识别鲁棒的可实现驱动节点集。
摘要：
由于生物网络的吸引子反映了稳定行为（例如细胞表型），因此识别能够驱动系统趋向其吸引子的控制干预措施（吸引子控制）在控制生物系统时具有特别重要的意义。通过节点状态覆盖将网络的反馈顶点集（FVS）驱动到与目标吸引子一致的状态，被证明可以迫使每条系统轨迹趋向目标吸引子，但在生物网络中，FVS通常大于实际可操作的范围。对生物网络FVS的一个子集进行外部控制被提出作为一种利用较少干预将网络驱动到其吸引子的策略；然而，该策略的有效性仅在一小部分生物网络的布尔模型上得到了验证。在此，我们将此分析扩展到生物启发的布尔网络集合。在这些模型上，我们使用三种结构指标——PRINCE传播、修改后的PRINCE传播和CheiRank——根据其预测的吸引子控制强度对FVS子集进行排序。我们使用三种动态测度：To Control、Away Control和逻辑影响域来验证这些排序的准确性。我们还在有效图上计算传播指标，这些图将每个布尔模型的功能信息纳入边权重。虽然这些额外信息增加了结构指标的预测能力，但我们发现相对于未加权网络的增加是有限的。传播指标结合FVS可用于通过模拟生物网络中普遍存在的动态来识别可实现的驱动节点集。这种方法仅使用网络的结构，并且驱动节点集被证明对特定的动态模型具有鲁棒性。
英文摘要：
Because the attractors of biological networks reflect stable behaviors (e.g., cell phenotypes), identifying control interventions that can drive a system towards its attractors (attractor control) is of particular relevance when controlling biological systems. Driving a network's feedback vertex set (FVS) by node-state override into a state consistent with a target attractor is proven to force every system trajectory to the target attractor, but in biological networks, the FVS is typically larger than can be realistically manipulated. External control of a subset of a biological network's FVS was proposed as a strategy to drive the network to its attractors utilizing fewer interventions; however, the effectiveness of this strategy was only demonstrated on a small set of Boolean models of biological networks. Here, we extend this analysis to ensembles of biologically-inspired Boolean networks. On these models, we use three structural metrics -- PRINCE propagation, modified PRINCE propagation, and CheiRank -- to rank FVS subsets by their predicted attractor control strength. We validate the accuracy of these rankings using three dynamical measures: To Control, Away Control, and logical domain of influence. We also calculate the propagation metrics on effective graphs, which incorporate each Boolean model's functional information into edge weights. While this additional information increases the predicting power of structural metrics, we find that the increase with respect to the unweighted network is limited. The propagation metrics in conjunction with the FVS can be used to identify realizable driver node sets by emulating the dynamics that are prevalent in biological networks. This approach only uses the network's structure, and the driver sets are shown to be robust to the specific dynamical model.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
