Predicting response to therapy in the multiverse

Predicting response to therapy remains a critical unmet need in precision oncology, as early identification of responders and non-responders enables timely therapy escalation or de-escalation, minimizes exposure to ineffective or toxic regimens, and supports implementation of adaptive treatment strategies. Our lab develops integrated frameworks to predict therapeutic response across the “multiverse”—in realistic preclinical models of breast cancer (co-clinical trials), prospective clinical trials, and increasingly through virtual imaging trials. Quantitative PET and MR imaging serve as the foundation for deriving clinically relevant insights into tumor biology and the tumor microenvironment. We combine co-clinical imaging with advanced image analytics and multi-scale biological data (including genomics and pathology) to characterize imaging biomarkers and build predictive models of treatment efficacy.  We have further developed technologies to enable virtual imaging trials to further support this effort, enabling controlled evaluation of imaging protocols and predictive biomarkers across diverse scenarios.  The effort is driven by a multi-disciplinary team of investigators spanning the domains of preclinical and clinical PET and MR imaging, imaging informatics, imaging physics, computational biology, artificial intelligence/machine learning, molecular oncology, and animal modeling.   The presentation will include examples of recent research activities and opportunities in this domain.