Beyond RECIST: mathematical modeling and Bayesian inference reveals immune parameters predict site specific response in metastatic breast cancer

Immunotherapies that target the host immune system to mount effective responses to cancer hold great promise. Overcoming patient- and organ-specific heterogeneity remain a significant challenge. In order to quantify individual patient responses to treatment, we fit a tumor-immune mathematical model to patient and site-specific dynamics of response to combination treatment (nivolumab + ipilimumab + entinostat) using tumor data (RECIST) coupled with immune markers measured by imaging mass cytometry. Bayesian parameter inference of the site-specific patient responses reveals that only immunosuppression parameters can explain response; parameters controlling cytotoxicity are not predictive. Via the fits of many tumors, we quantify the variability in tumor-immune dynamics across patients and tissues and reveal controllable parameter regimes. We go on to show that through posterior parameter sampling and simulation, we are able to use our model to extrapolate beyond the data and predict the probability of response in virtual metastatic tumors in patients for which we have no data at that site, thus overcoming the limitations of a small clinical trial to enable the analysis of a large virtual patient and virtual tumor cohort undergoing combination treatment.

Beyond RECIST: mathematical modeling and Bayesian inference reveals immune parameters predict site specific response in metastatic breast cancer

Monday, July 14 at 6:00pm

Jesse Kreger

"Beyond RECIST: mathematical modeling and Bayesian inference reveals immune parameters predict site specific response in metastatic breast cancer"

#SMB2025 Follow