Aim of this work package is to create virtual portions of tumors and micro-metastasis for each patient and to simulate in silico treatments on the basis of the individual characteristics of the cancer cells. Plan is to use mechanistic and multiscale mathematical models of cancer cells response to therapy and exploit individual high throughput imaging and genomic data to perform inference on model components and parameters. The tailoring to individual patients is based on the estimation of those parameters in order to truly account for biological and clinical heterogeneity.
This is the workflow to run a personalized simulation of drug response of a breast tumor portion: simulation and validation. Tumor screening data at week 0 are used, to initialize and parameterize the model, while MRI examinations at weeks 1 and 12 are used to validate the simulated outcomes (box A). After initializing and parameterizing the mathematical model using patient data and public data in literature (boxes B and C), we simulate the drug schedule used in the clinical trial for each patient (box D). For that, we increase the amount of the corresponding drugs (FEC100 +/-bevacizumab) in the blood at the time points of administration according to the dosage. Each computer simulation then runs a complete cycle of 12 weeks of the spatio-temporal dynamics of the considered tumor portion under the effect of the applied drugs (box E). Computational details and a link to the open-source computer code can be found in section S3 of SM. Simulated outcomes include the spatial distribution of cells, blood vessels and all considered molecules at any time between the start of the therapy and week 12 and in particular at the end of the study period (box F). To validate a simulation, we compare the simulated outcomes with the actual ones (box H). For that we use clinical tumor volumes and apparent diffusion coefficients (ADC) calculated from DW-MR imaging at week 1 and 12 (box G).
