Fig. 3: Dynamical mutual information and comparison with the previous approaches.

We used all the 13 different stimulus conditions to calculate the mutual information. a The maximum dMI increases with time and is diminished when timepoints are randomly shuffled. To be consistent with other MI measures, all y-axis of the maximum dMI is labeled as “Maximum MI.” b The timepoint method¹⁸ and the vector method²⁷ do not capture information in the full signaling trajectory. The vector method suggests that mutual information decreases after around 10 timepoints, while the trajectory ensembles retain substantial differences, and is independent of the ordering of timepoints. c The decoding-based method²⁸ (using the first ten principle components) provides a lower bound on mutual information (y-axis is “MI” without “Maximum”) and does not reveal how mutual information accumulates over time. The random permutation of timepoints does not dramatically affect the saturated value. Source data are provided as a Source Data file.