The advancements in computational biology in tandem with the explosion of machine learning (ML) have the potential to revolutionize quantitative and personalized medicine. In this framework, a Deep Operator Network (DeepONet) enhanced computational framework is presented, which enables informed simulations of cancer growth. A continuum mechanics-based, biphasic tumor growth model (TGM) is developed that simulates the tumor as well as the immune response within the tumor microenvironment (TME). A Bayesian inference framework is then customized in order to calculate the posterior distributions of critical model parameters using experimental data. Given the computational complexity of solving the TGM consisting of multiple coupled nonlinear partial differential equations (PDEs), we integrate a DeepONet surrogate model to significantly reduce computational costs while preserving the accuracy of the original model. This approach enhances the identification of model parameters based on experimental data, underscoring the importance of Bayesian strategies in cancer modeling. The proposed framework offers a robust and adaptable approach to tumor growth simulations, paving the way for more precise, data-driven predictions in cancer research and treatment.
Sotiropoulos et al. (2026) studied this question.