Spatial agent-based modeling and interpretable machine learning identify determinants of combination-therapy response in HER2-heterogeneous breast cancer

Two documented clinical failures match this mechanism, or a single Phase 3 failure is on record.

Abstract excerpt

HER2 heterogeneity and reversible phenotypic plasticity play a central role in breast cancer progression and therapeutic resistance, yet how their interaction shapes treatment response remains poorly understood. HER2-positive and HER2-negative tumor cell states can dynamically interconvert, enabling compensatory population shifts that undermine monotherapies targeting a single phenotype. Because stochastic lineage effects and local cell interactions are averaged out in mean-field population-level ODE models, we develop a spatially resolved agent-based model (ABM) of heterogeneous tumor growth. We consider paclitaxel, which is modeled to preferentially suppress HER2-positive proliferation, and Notch inhibition, which targets HER2-negative populations and alters phenotypic composition. Starting from single-cell lineages, we assess the ABM against theoretical predictions from a population-level switching model and against single-cell-derived experimental measurements, showing consistency with early lineage dynamics and long-term phenotypic equilibria. Simulation results show that monotherapies induce compensatory phenotypic shifts and spatial reorganization that permit tumor persistence. In contrast, combination therapy simultaneously targeting HER2-positive and HER2-negative populations disrupts phenotypic replenishment, fragments spatial structure, and can achieve sustained tumor control across a range of simulated tumor regimes and treatment strengths. Importantly, in matched tumors with identical geometry, total burden, and phenotype counts, peripheral enrichment of HER2-negative cells increased post-treatment escape under paclitaxel, whereas this spatial effect was nearly eliminated by combination therapy. To quantify robustness across heterogeneous tumor parameter regimes, we pair the ABM with an interpretable Random Forest surrogate. Using only pre-treatment and early-trajectory features, the surrogate discriminates sustained control from persistence across held-out simulated parameter regimes and identifies growth-rate asymmetries as dominant drivers of resistance. Together, this integrated mechanistic and data-driven framework clarifies how HER2-mediated plasticity, spatial organization, and competitive growth dynamics shape therapy resistance and provides a scalable approach for analyzing simulated treatment responses across heterogeneous tumor regimes.