Cancer is a dynamic and heterogeneous disease and these two properties underpin cancer’s uncanny ability to develop resistance to targeted therapies. To overcome the almost inevitable emergence of drug resistance it is necessary to explore the (protein) network level dynamics elicited by drug treatment (dynamics) and how these dynamic responses vary across a wide range of network conditions (heterogeneity). Estrogen Receptor (ER) positive breast cancer is a useful system with which to explore these concepts as it is well characterised but still frequently presents resistance to its current standard treatment (estrogen deprivation and Cyclin Dependent Kinase 4/6 (CDK4/6) inhibition).
In this investigation we began with the construction of a mechanistic and dynamic mathematical model that describes the multi-pathway protein network surrounding CDK4/6 and the Estrogen Receptor (ER). Then, using a novel ensemble modelling approach, we explored the range of potential behaviours exhibited by this network in response to both CDK4/6 and ER inhibition, by comprehensively sampling the constituent parameter values and protein concentrations within several orders of magnitude.
Our investigations reveal that the CDK4/6 network possesses numerous conditions that facilitate resistance and that these conditions are widespread across parameter values and protein concentrations. Further analysis shows resistance does not appear to be dependent on any particular parameter values, or protein concentrations. To explore the possibility that there still might be parameters/concentrations that drive resistance in a relative manner, we performed sensitivity analyses across the entire range of network conditions. We were able to demonstrate that there are in fact, distinct patterns of protein contribution to resistance and many of these patterns parallel mechanisms of resistance seen in the literature. This investigation demonstrates that through the systematic analysis of network behaviour across a broad range of conditions, it is possible to identify network patterns that could be exploited to rationally select targets for combination therapy that addresses both intra-tumoral and inter-patient heterogeneity.