Identifiability and sensitivity analysis for the CD8 T cell lineage.

A. Sensitivity of relative cluster sizes () with respect to differentiation rates Q ij . Each plot contains d = 8 curves . The primary effects are highlighted in black ( i = k ) and red ( j = k ), while the secondary effects are shown in gray. The y -axes are shown on a square-root-scale. B. Sensitivity of total population size ( Y ( t )) with respect to differentiation rates Q ij . The curves correspond to . C. Practical identifiability scores of differentiation rates Q ij . The heatmap shows the correlation coefficient between the ground truth value of Q ij , and the estimated value. The stars indicate the levels of statistical significance ( p < 0.1, * p < 0.05, ** p < 0.01, *** p < 0.001) based on a range of 51 ground truth Q ij values. D. A single pseudo-dataset is simulated with model IV, using parameters estimated from the true data. Model IV is then fit to the pseudo-data, and for each pair of populations ( i , j ) we show the estimate versus the ground truth value of Q ij . E. Model identifiability. Data was simulated with and fit to each of the four models, resulting in 16 model fits. For each simulated dataset, the four model fits are compared with model weights (shown in color). High diagonal model weights indicate that the ground-truth model is correctly identified. Shown is the median of 3 simulations for each model. (PDF)