Abstract:We are interested in the relationships among network topology, robustness, and identifiability, and their implications in improving network reconstruction. We used three different types of artificial gene networks (AGNs) with distinct topologies: topologies random (RND), scale‐free (SF), and small‐world (SW), to investigate their robustness and identifiability. The robustness of a network is represented by structural reachability (existence of pathways between two nodes) and dynamic reachability (response on one node upon perturbation on another node). The identifiability of the network edges is assessed in silico with an established reverse‐engineering algorithm. We found that (1) structural reachability does not always lead to dynamic reachability; (2) network robustness is high and identifiability is low in all surveyed AGNs; (3) robustness is more sensitive to network topologies than is identifiability. We also devised a method for network dissection in which three subnets (set of alternative pathways or feedbacks, referred to as pathnet) are related to each node pair. This method allows us to identify the fine structural features underlying the distinct behaviors of the networks. For example, pathnet of the edge tail negatively contributes to the edge identifiability, and it is likely that extra perturbation at this pathnet would improve edge identifiability. We provide a case study to prove that double perturbations decrease the edge robustness and increase structural identifiability with a T helper cell–differentiation network model.

Nonlinear network identifiability: The static case

Nonlinear Network Identifiability with Full Excitations

Local Identifiability of Networks with Nonlinear Node Dynamics

On Necessary and Sufficient Conditions for Identifiability and Identification of Switching Dynamical Networks

Nonlinear identifiability of directed acyclic graphs with partial excitation and measurement

Identifiability and Identification of Switching Dynamical Networks: A Data-Based Approach

Robust Network Tomography: K-Identifiability and Monitor Assignment

Single module identifiability in linear dynamic networks

Identifiability in Dynamic Acyclic Networks with Partial Excitations and Measurements

Network Identification for Diffusively-Coupled Systems with Minimal Time Complexity

A nonlinear graph-based theory for dynamical network observability

A scalable multi-step least squares method for network identification with unknown disturbance topology

Identification For Control Based on Neural Networks: Approximately Linearizable Models

Global Robustness and Identifiability of Random, Scale‐Free, and Small‐World Networks

Functional Network Method for the Identification of Nonlinear Systems

About identifiability and observability for a class of dynamical systems

Identifiability in Continuous Lyapunov Models

Identifiability of catenary and directed-cycle linear compartmental models

Extending identifiability results from isolated networks to embedded networks

Bayesian topology identification of linear dynamic networks

The informative enough property of the data set in a Networked Control System