Predicting the response of triple negative breast cancer to neoadjuvant systemic therapy via biology-based modeling and habitat analysis

Despite being the standard-of-care treatment, neoadjuvant therapy (NAT) attains a complete response only in approximately half of the patients with triple negative breast cancer. Thus, methods to predict and optimize patient response to NAT are needed. Previously, we employed patient-specific MRI data to calibrate a biology-based mathematical model that describes cell movement, proliferation, and death due to drug at the tumor level and cell proliferation at an image voxel level. We now extend our approach by using MRI data to group voxels into "habitats" whereby tumor cells of a habitat share the same proliferation. With this approach, we now calibrate habitat-informed proliferation rates for each habitat rather than local proliferation rates. When comparing error in tumor cell number and volume at the time of calibration, the local calibration has significantly (p < 0.05) lower error than the habitat-informed calibration. However, the habitat-informed predictions of a future timepoint have significantly lower error than the local predictions. Compared to the local calibration, the habitat-informed calibration also requires fewer parameters, reducing the calibration time by a factor of 17. These results suggest that a habitat-informed calibration can provide both accurate and efficient predictions of breast cancer response to NAT.