Approximate Bayesian Computation Applied to Model Selection and Parameter Calibration in Cell Proliferation

Abstract

Approximate Bayesian Computation is used in this work for the selection and calibration of cell proliferation models. Four competing models based on ordinary differential equations are analyzed, by using the measurements of the proliferation of DU-145 prostate cancer viable cells during seven days. The selection criterion of the ABC algorithm is based on the Euclidean distance between the model prediction and the experimental observations. The Richards Model and the Generalized Logistic Model were selected by the ABC algorithm used in this work, providing accurate estimates of the evolution of the number of viable cells. Bayes factor revealed that there was no evidence in favor of any of these two selected models.

Full document