Data-Driven Modeling of Macrophage Polarization Dynamics and Cytokine Feedback to Predict Immune Modulation Strategies

Macrophages are highly plastic immune cells that dynamically polarize along a spectrum between pro-inflammatory (M1-like) and pro-healing (M2-like) phenotypes in response to environmental cues. In the tumor microenvironment and during chronic inflammation, dysregulated polarization and efferocytosis drive pathological immune responses and disease progression. To elucidate the regulatory principles governing macrophage state transitions, we develop a mechanistic, data-driven model that integrates macrophage polarization with key cytokine networks involving TNF-α, IL-6, IL-10, and IFN-γ. The system is formulated as nonlinear ordinary differential equations (ODEs) representing resting, M1-like, and M2-like macrophages and their cytokine mediators. Parameters are optimized using bulk RNA-seq time-series data from LPS+IFN-γ stimulation experiments and validated with single-cell RNA-seq–derived polarization trajectories and M2→M1 repolarization datasets. Bayesian model selection and averaging are employed to compare alternative network topologies and quantify uncertainty, yielding a robust predictive framework for macrophage–cytokine crosstalk. Sensitivity and perturbation analyses identify key parameter, such as IL-10–mediated feedback and efferocytosis efficiency, that control transitions between inflammatory and resolving immune states. The resulting model provides a quantitative foundation for forecasting outcomes of chronic inflammation and for designing therapeutic interventions that modulate macrophage polarization. This approach offers a principled path toward TAM-reprogramming strategies in cancer and other inflammatory diseases by bridging omic data, mechanistic modeling, and predictive immunomodulation.