Abstract P1-08-20: In silico analysis of a novel mathematical model integrating in vitro and in vivo imaging data reveals driving mechanisms of breast cancer response to NAT for personalized tumor forecasting

Guillermo Lorenzo,Angela M. Jarrett,Christian T. Meyer,Darren R. Tyson,Vito Quaranta,Thomas E. Yankeelov

DOI: https://doi.org/10.1158/1538-7445.sabcs21-p1-08-20

IF: 11.2

2022-02-15

Cancer Research

Abstract:Abstract The early assessment of neoadjuvant therapy (NAT) response in triple-negative breast cancer (TNBC) would enable a treating oncologist to adjust a therapeutic plan of a non-responding patient, and thereby enhance outcomes while preventing unnecessary toxicities. To address this challenge, we propose leveraging personalized, in silico forecasts of tumor response to therapeutic regimens via a mechanistic mathematical model calibrated with patient-specific longitudinal magnetic resonance imaging (MRI) data acquired early during NAT. Here, we focus on identifying the driving mechanisms involved in the model formulation through a global sensitivity analysis. Our model describes the dynamics of TNBC cell density as a combination of mobility, which is formulated as a diffusion process constrained by local tumor-induced mechanical stress, and net proliferation, which is represented with a logistic term. To model TNBC response to NAT drug combinations, we adjust the tumor cell proliferation rate with a recent pharmacodynamic model, MuSyC, which also accounts for synergy of potency and efficacy. Tumor cell density is estimated from diffusion-weighted MRI data, while tissue mechanical properties are defined from segmented contrast-enhanced T1-weighted MRI data. To model the heterogeneous intratumoral delivery of drugs, we use perfusion maps estimated from dynamic contrast-enhanced MRI data. NAT drug pharmacokinetics are approximated with a linear model, which reasonably represents their temporal decay during NAT. Sobol’s method is used for the global sensitivity analysis of the NAT response of two different tumors (one well-perfused and one poorly-perfused) on a 3D, tissue-scale domain. This allows us to assess the total effect (ST) of each model parameter on tumor volume and global cellularity. Here, we focus on two standard NAT regimens: doxorubicin plus cyclophosphamide, and paclitaxel plus carboplatin. The parameter space is constructed by integrating three approaches. First, we use prior patient-specific in silico estimates of tumor cell mobility and proliferation. Second, we experimentally constrain the parameters accounting for potential synergistic drug activity by using time-resolved, high-throughput, automated microscopy assays that capture drug-induced changes in proliferation rates of various TNBC lines (HCC1143, SUM149, MDAMB231, and MDAMB468; perfosfamide was used in lieu of the pro-drug cyclophosphamide). Third, we scale the resulting in vitro parameter ranges to be clinically-relevant through an in silico study with our mechanistic model. Our results show that out of the 15 model parameters considered in the sensitivity analysis only a minority exhibited a driving role (ST > 0.1) in representing the dynamics of TNBC response to NAT, namely: the baseline tumor cell proliferation rate along with the effect and ratio of peak concentration to EC50 of doxorubicin, paclitaxel, and carboplatin. The other parameters have a marginal effect and can thus be fixed to any value within the parameter space. We select these constant parameters such that they contribute to simplifying the model formulation, leading to a surrogate reduced model. We further show that the reduced and the original model produce distributions of tumor volume and global cellularity that are in remarkable agreement both qualitatively (Dice similarity coefficient > 0.90 and > 0.94, respectively) and parameter-wise (concordance correlation coefficient > 0.85 and > 0.89, respectively). Thus, we conclude that our reduced model constitutes a feasible surrogate for future clinical calibration-forecasting studies, thereby facilitating personalized NAT response forecasting with patient-specific imaging datasets acquired in vivo. Citation Format: Guillermo Lorenzo, Angela M. Jarrett, Christian T. Meyer, Darren R. Tyson, Vito Quaranta, Thomas E. Yankeelov. In silico analysis of a novel mathematical model integrating in vitro and in vivo imaging data reveals driving mechanisms of breast cancer response to NAT for personalized tumor forecasting [abstract]. In: Proceedings of the 2021 San Antonio Breast Cancer Symposium; 2021 Dec 7-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2022;82(4 Suppl):Abstract nr P1-08-20.

oncology