Fig 3. Statistics for per-minute changes in the experimental time series.

(A-D) Change per minute for the experimental time series at stripe-center (A, B; data of Fig 1B and 1C, respectively); the anterior stripe edge (C; data of Fig 1D); and the posterior stripe edge (D; data of Fig 1E). Change per minute is calculated as the difference in number of nascent transcripts divided by the difference in time for each pair of successive data points. This is the net change (number of initiating transcripts–number of terminating transcripts per minute), i.e. intervals with more initiation than termination have positive change. Pink indicates high initiation intervals (gains of 10 or more nascent transcripts per minute); blue indicates low to mid initiation intervals (below 10 additions per minute). The threshold of 10 is for illustrative purposes, it is not used in the statistical analysis of the experimental data or model results. (E-H) Corresponding histograms of the per-minute changes (E ↔ A; F ↔ B; G ↔ C; H ↔ D), labeled at lower class limits. (E, F) At stripe-center, the distribution of change rates is broad, ranging from a maximum net addition of 33.5 transcripts in a minute (A, minute 15) to a maximum net loss of 29 transcripts in a minute (B, minute 20). While low net-change minutes (near zero addition) are most common, mid- to high-change minutes are also well populated. Approximately 55% of per-minute changes are positive. (G, H) At the stripe edges, with higher concentrations of the Gt and Kr repressors, the distribution of rates becomes narrower: to the anterior (G), at 38%EL, the histogram is well-filled for low to medium rates, but shows less frequent high intensity intervals than at stripe-center; to the posterior (H), at 46%EL, while high-intensity minutes can still be accessed (e.g. D, minute 14), per-minute change is predominantly low to mid, and the distribution is more uneven than at 38%EL, with low to zero frequency for many of the rates. (I-J) Autocorrelation plots for the corresponding per-minute changes (I ↔ A; J ↔ B; K ↔ C; L ↔ D). The per-minute changes show no significant autocorrelation at stripe-center or the posterior edge (I, J, L; red lines are 5% significance limits, zero ± t*SE, see text), and autocorrelation at 1 minute lag for the data at the anterior edge of the stripe (K).