Abstract:
What would you do if you knew everything about a patient’s tumor? How should a physician make a treatment decision with multi-omics and imaging data at single-cell resolution?
We sought to answer these questions by constructing a deterministic, 4-D, multi-scale mathematical model of tumor biology called TumorScope that incorporates multi-omics data, imaging data, pathology, and patient characteristics to simulate tumor biology in individual patient tumors.
We have validated this approach in breast cancer by comparing TumorScope simulations vs actual tumor response to neoadjuvant chemotherapy in over 400 patients, with a median volumetric error < 5% at time of surgery (several months later). Furthermore, using TumorScope, we predict patient response to multiple chemotherapies as a tool for physician decision support. The mechanistic nature of TumorScope and its basis in biology ensure predictions are fully traceable, as opposed to “black-box” AI, and can be logically checked by treating physicians.
