Figure 9.
Parameter distributions of consecutive predictions following weekly data assimilation. (a) Prediction of bulk cell count on a replicate treated with 50 nM of doxorubicin. (b) Prediction of cell count on pixel (6,8). (c) Bootstrapped parameter distributions for model predictions on pixel (6,8), showing the shift in the mean value and variance of the post-treatment parameters between each weekly timeframe. The distribution for the first weekly segment of parameter fs is indicated by a star as it is so narrow and may be difficult to see. The changes in the width of the parameter distributions is representative of changes in the uncertainty in the parameter value, which emanate from the quality of the model fit in the preceding weeks as well as from the wide variety of parameter values in similar pixels in the training set. The changes in the mean value point toward changes in the overall dynamics identified by the model as more data is progressively assimilated on a weekly basis.