Cancer therapy and oncology has entered a new exciting era of targeted therapy and personalised patient treatment, but resistance and tumour heterogeneity represent a significant hurdle for realizing the clinical impact of these discoveries. Overcoming this hurdle requires an ability to quantitatively describe heterogeneous tumour cell populations and their dynamic response to treatment over time. Computational modelling in close conjunction with experimental validation represents an attractive avenue towards evaluating which drugs, combinations, and schedules are best for a given patient. While it is unethical and too time-consuming to test all possible drug combinations and dosing schedules in pre-clinical or clinical studies (and therefore only a limited clinical experimentation can be performed), computational modelling can, in principle, be used to narrow the set of possibilities to identify the combinations and schedules that maximize patient survival. Here, we apply a systems approach that combines predictive network modelling with experimentation to discover and prioritize synergistic drug combinations for triple-negative breast cancer (TNBC), an aggressive breast cancer subtype.
We have previously showed that the non-receptor tyrosine kinase PYK2 is a downstream effector of both EGFR and c-Met and their crosstalk signalling in basal-like TNBC. Using network-level data, we develop and train a dynamic mathematical model of the integrated EGFR-PYK2-c-Met signalling network. Model simulations of numerous pair-wise target combinations predict that co-targeting EGFR/PYK2 induce the most potent synergistic effect in TNBC cell lines, followed by co-targeting EGFR/c-Met, predictions that we subsequently validated experimentally. Model-based analysis reveals that the synergy brought about by co-targeting EGFR/PYK2 was linked to the ultrasensitive switch-like response at the signalling and cell proliferation levels. Furthermore, analysis of TNBC patient clinical data and patient-specific model simulations enable systematic stratification of these patients into distinct subgroups with predicted high and low susceptibility to EGFR/PYK2 co-targeting. Our findings demonstrate integrative network modelling as a promising approach to rationalize target and drug combinations for improved treatment of TNBC. We believe that this approach could be widely applicable to other cancer types given the appropriate data for model development, training and validation.