Towards personalized computer simulation of breast cancer treatment: a multi-scale pharmacokinetic and pharmacodynamic model informed by multi-type patient data

Abstract

Mathematical modeling and simulation have emerged as a potentially powerful, time and cost effective approach to personalized cancer treatment. The usefulness of mechanistic models to disentangle complex multi-scale cancer processes such as treatment response has been widely acknowledged. However, a major barrier for multi-scale models to predict the outcomes of therapeutic regimens in a particular patient lies in their initialization and parameterization which need to reflect individual cancer characteristics accurately. In this study we use multi-type routinely acquired measurements on a single breast tumor, including histopathology, magnetic resonance imaging, and molecular profiling to personalize parts of a complex multi-scale model of breast cancer treated with chemotherapeutic and anti-angiogenic agents. We model the dynamics of drugs in tissue (pharmacokinetics) and the corresponding effects on their targets (pharmacodynamics). We developed a open-source computer program that simulates cross-sections of tumors under 12-week 1 therapy regimes and use it to individually reproduce and elucidate treatment outcomes of four patients. For two of the tumors that did not respond to therapy, we used model simulations to suggest alternative regimes, depending on their individual characteristics, with improved outcomes. We found that more frequent doses of chemothereapy reduce tumor burden in a low proliferative tumor while lower doses of anti-angiogenic agents improve drug penetration in a poorly perfused tumor. In addition to bridge multi-type clinical data to shed light on individual treatment outcomes, our approach identified a few tumor-related aspects that need to be clinically portraited better to allow for future model-driven personalized cancer therapy.

ABC of the future

“…Personalised cancer treatment is an application area where simulator-based inference has a lot of potential, given the constantly improving biological generative models for the evolution of the disease under treatment (Sottoriva & Tavaré, 2010;Kozłowska et al, 2018;Lai et al, 2019;Lai et al, 2021). Realistic biological generative tumor models that are built up from the molecular and cellular processes result typically in a level of complexity that renders analytical solutions infeasible.…”

“…The cancer simulator we use is a multi-scale pharmacokinetic and pharmacodynamic model describing the response of a cross-section of breast tumor tissue to a combination of chemotheraupetic and anti-angiogenic agents (Lai et al, 2019). Mathematically, it consists of a hybrid cellular automaton model (Ribba et al, 2004;Alarcón et al, 2010) that couples stochastic and discrete model formalisms with deterministic and continuous components accounting for biological processes at different spatio-temporal scales; see Figure 3.…”

“…Mathematically, it consists of a hybrid cellular automaton model (Ribba et al, 2004;Alarcón et al, 2010) that couples stochastic and discrete model formalisms with deterministic and continuous components accounting for biological processes at different spatio-temporal scales; see Figure 3. There are multiple (Lai et al, 2019) for full model details and its patient-specific parameterisation parameters in the model influencing the evolution of cancer cells, stromal cells and tumour vessels, which are all modeled as individual agents. Many of the parameters can be inferred and fixed via various means but some of the unknown parameters have to be inferred based on the reaction of the tumour to the treatment.…”

“…The right column shows the different model formalisms used for each of the model modules. See(Lai et al, 2019) for full model details and its patient-specific parameterisation…”

ABC of the Future

“…Personalised cancer treatment is an application area where simulator-based inference has a lot of potential, given the constantly improving biological generative models for the evolution of the disease under treatment [Sottoriva and Tavaré, 2010, Koz lowska et al, 2018, Lai et al, 2019. Realistic biological generative tumor models that are built up from the molecular and cellular processes result typically in a level of complexity that renders analytical solutions infeasible.…”