A workflow of the Meta Dynamic Network (MDN) modelling pipeline.

Process flow diagram showing the steps of the MDN modelling process.

Network Schematic of the Early Cell Cycle (ECC) Network.

Network schematic showing the detailed biochemical reactions that regulate the initiation of the G1 phase and transition through to the S phase of the cell cycle.

The ECC network robustly facilitates resistance-associated protein dynamics.

Clustered heatmap of a representative sub group of the time-course responses of phosphorylated Rb (pRb) to CDK4/6 and ER inhibition, across parametric variation. B) The frequency of each dynamic category for a selection of active protein forms across 100,000 model instances. C) The frequency of model instances displaying simultaneous sensitivity (blue), or resistance in at least one of the key output proteins (red).

Protein interactions are a stronger driver of adaptive-resistance dynamics than protein expression.

A) Overview of the computational pipeline undertaken to compare the effects of parametric variation with state variable variation. B) The frequency of which state variable (IC) variation (left) induces resistance, versus parameter variation (right). C) The number of unique dynamic categories produced by state variable variation (left), versus parametric variation (right).

MDN analysis identifies core sub-networks that facilitate resistance.

A) Overview of the process undertaken to investigate the relationship between parameter knockdown and resistance-dynamic features to identify resistance driving parameter signatures. B) Heatmap produced by hierarchical clustering of the parameters that contribute to rebounding pppRb, for model instances that display rebounding pppRb. C) Comparison of cluster-specific parameter rankings with the overall parameter rankings, where parameters are ranked by how frequently they contribute to rebounding pppRb. Clusters are aligned with overall ranking. The lower bar graph represents the correlation between the overall ranking and each cluster specific ranking.

Identification of subnetworks driving pppRb rebound-mediated resistance within the ECC network.

A) Cluster-specific parameter signatures overlaid with the ECC network to highlight subnetworks that drive resistance through rebound of pppRb. Only the five largest clusters (out of 9) are displayed. B) Parameter knockdowns for the top-scoring parameters in three select clusters that display divergent resistance driving parameter signatures. Black represents treatment with CDK4/6 and ER inhibitors alone, decreasingly-red lines represent the addition of an increasingly potent parameter knockdown.

Validation of MDN-based predictions.

A) Single cell time-course responses to CDK4/6 inhibitors (obtained from Yang et al. [55]). Red lines are resistance-associated dynamics, blue lines are sensitivity-associated dynamics, and grey lines are no/limited response dynamics. B) Frequency of each dynamic category in panel A. C) Plotting the dynamic responses of CDK46cycD and CDK2cycE from a random selection of model instances generated with the MDN pipeline (parametric variation). D) Comparison of single-cell experimental data and MDN predictions, looking at the differences between the frequency of CDK4/6 and CDK2 dynamics.