Skip to main content
2023 BMES Annual Meeting Main banner
Icon Legend
This session is not in your schedule.
This session is in your schedule. Click again to remove it.

Cancer Technologies

Computational Modeling of Cancer and Systems Biology

Clinical combination therapy optimization through personalized mechanistic models

Thursday, October 12, 2023
12:45 PM - 1:00 PM PDT
Location: 341 (Center, Level 3)
Introduction:: As discoveries related to cancer genomics have rapidly increased, patients diagnosed with high-risk cancers are subjected to largely empiric and intensive combination regimens of chemotherapeutics and radiation therapy. Although effective, these regimens have substantial acute and long-term side effects. Traditional modeling of combination chemotherapies involves performing simple addition of the independent effects of individual chemotherapies, which are available in literature as population values without guidance for adapting them for specific patients. Radiation therapy doses are traditionally prescribed uniformly for a particular subtype and stage of cancer, despite high patient-to-patient variability in responses[1]. To personalize treatment regimens, methods that can accurately predict combination response are needed. Here we developed a mechanistic modeling framework that simulates tumor cell growth and response to combination therapy. In this framework, we mechanistically model disease-implicated cell signaling networks and the effects of therapeutics on these networks based on empirically determined values. Personalized response predictions are obtained by constraining initial model states based on patient-specific molecular profiles. Additionally, agent-based modeling (ABM) is used to simulate the effects of patient-specific intra-tumor heterogeneity on responses.

Materials and Methods:: MiRNA profiles were collected from tumor tissue from 36 Wilms Tumor (WT) patients who were treated with combination therapy according to the European SIOP protocol[2]. Tumor volume data before and after chemotherapy are available for 16 of these patients. We performed gene ontology enrichment of differentially expressed miRNAs across WT subtypes to identify disease- and subtype-specific pathways. We created cell signaling models of WT by hybridizing models of implicated pathways, using ordinary differential equation models to represent empirically characterized pathways and Boolean networks to represent known logical relationships between pathway species in the absence of empirically determined kinetic data. Tumor cell death and cell growth probabilities were modeled as the rates of activation of P53-mediated cell death and cell proliferation signaling pathways across ensemble simulation runs covering the space of initial states possible for the model network. To predict response to a specific chemotherapy, we constrained the initial state of model species known to be targeted by the drug. To predict the response to radiation therapy, we integrated the linear-quadratic model, which estimates cell survival post radiation-induced DNA damage as a function of radiation dose, into our pathway model of P53-mediated DNA damage repair. Ensemble simulation runs were then performed for varied combinations of chemotherapies and radiation dosage. To predict effects of intra-tumoral heterogeneity on response predictions, we constructed an ABM using cell fate probabilities estimated from our pathways model and patient histology data, and then simulated responses to various combination therapy regimens over a one-month period.

Results, Conclusions, and Discussions:: Results and Discussion: The resulting cell signaling model incorporated P53-mediated cell cycle and EGF-mediated PI3K and MAPK signaling pathways and notably included P53, IGF2, and MYCN, which are key prognostic markers in high and intermediate risk WT. As a mechanistic rather than data-driven model, predictions for each patient and control were estimated over an ensemble of simulations covering the initial state space possible from the unconstrained species within the model and without the need of fitting parameters from training data. Model analysis revealed high variability in WT patient responses to different therapy combinations for both chemotherapy and radiation therapy. Only a subset of therapy combinations were predicted effective in killing blastemal cells, which represent the least differentiated cells in a WT and are over-represented in higher risk WTs. Simulating application of therapy combinations used on patients in clinic resulted in predicted tumor cell kill probabilities that correlated with clinically observed tumor volume changes. Simulating patient tumor histological compositions via ABM resulted in predicted responses that correlated with clinically determined WT subtype risk.


Conclusions: We demonstrate a novel predictive tool with capacity to predict clinical outcomes with inclusion of just an individual patient’s molecular profile. Use of such a tool may allow oncologists to better identify optimal chemoradiotherapy regimens for individual patients.

Acknowledgements (Optional): :

References (Optional): : References:

(1) Zahid, ... Enderling. Int J Radiat Oncol Biol Phys. 2021 Nov 1;111(3):693-704.

(2) Wegert, ... Gessler. Cancer Cell 2015 Feb 9;27(2):298-311.