Professor University of Waterloo Waterloo, Ontario, Canada
Disclosure(s):
Mohammad Kohandel: No financial relationships to disclose
Recent advancements in cancer immunotherapy, particularly through the use of immune checkpoint inhibitors like anti-PD-1 antibodies, have transformed oncological treatments. Despite their efficacy, the pharmacodynamics of these inhibitors vary significantly among patients, underscoring the need for a robust methodology to interpret their dynamics effectively. This talk introduces an innovative proposal that leverages Quantitative Systems Pharmacology (QSP), machine learning (ML), and digital twin technologies to enhance the predictive modeling of immunotherapies. By integrating ML, this approach aims to improve the precision and adaptability of QSP models by leveraging extensive datasets to optimize parameters and simulate complex biological interactions. The integration of Universal Physics-Informed Neural Networks (UPINNs) with QSP models is a key innovation, allowing for the learning of unknown components of differential equations that govern the dynamics of drug interactions and disease progression. UPINNs help to enhance the fidelity and efficiency of these models, adapting in real-time to new data inputs. Moreover, the development of digital twins - virtual replicas of individual patients - enables personalized medicine by simulating different treatment scenarios and predicting patient-specific responses to therapies. This technology can assess sex-specific differences in immune responses, optimize the efficacy of combination therapies like IL12-Nivolumab, and evaluate the risk and progression of cytokine release syndrome induced by immunotherapies.
