Matlab code to calibrate a structured-PDE model to data from in vitro experiments

Matlab code for the calibration of a PDE model of evolutionary dynamics of a well-mixed population of aggressive breast cancer cells from in vitro data on MCF7-sh-WISP2 cell line and bootstrapping for uncertainty quantification. For more details, see the associated publication: "Evolutionary dynamics of glucose-deprived cancer cells: insights from experimentally-informed mathematical modelling", by L. Almeida, J. Denis, N. Ferrand, T. Lorenzi, A. Prunet, M. Sabbah, C. Villa (corresponding author, author of code), 2024. The manuscript is published in the Journal of the Royal Society Interface, https://doi.org/10.1098/rsif.2023.0587 

Methods

Experimental data on MCF7-sh-WISP2 cells are used to carry out model calibration through a likelihood-maximisation method. In summary, the optimal parameter set (OPS) is obtained, through an iterative process, by minimising the weighted sum of squared residuals, employing the average and standard deviation of summary statistics from the experimental data, exploiting the in-built Matlab function bayesopt, which is based on Bayesian Optimisation. At each iteration, the PIDE-ODE system that constitutes the model is solved numerically using the FTCS method. Uncertainty quantification of the OPS was carried out by means of a bootstrapping algorithm, based on random sampling of data with replacement and particularly suited when only a few data are available. See the Supplementary Material of the associated publication for more details.