Abstract
We propose a system of ordinary differential equations to model the mouse immune response of two key immune cell types (CD4+ and CD8+ cells) to an established triple-negative breast cancer tumor while being treated with immunotherapy drugs of anti-PD-1 and anti-CTLA-4 immune checkpoint inhibitors. The model incorporates longitudinal positron emission tomography image data from a series of experiments where immunotherapy treatment was given in combination or separately. Control data optimization estimates the immune-tumor response of a general mouse burdened with breast cancer. Collaborative input designated the location of treatment effects that were further parameterized. The results indicate quantifiable differences in parameter values that differentiate immunotherapy responder and nonresponder groups. Treatment parameters are first determined from single and then from combination immunotherapy data. Structural identifiability is used to classify the identifiability of the parameters, while Sobol sensitivity analysis is employed to narrow the key treatment interactions of the model. From the constrained treatment model, we can accurately predict tumor volume changes for most treatment data, which strengthens our methodology while highlighting key interactions.