Fig. 5: Accuracy and robustness of the analysis and modeling pipeline. | Nature Communications

Fig. 5: Accuracy and robustness of the analysis and modeling pipeline.

From: Stochastic pausing at latent HIV-1 promoters generates transcriptional bursting

Fig. 5

A Accuracy and robustness of the deconvolution method. Left panels: simulation of short movies for an artificial set of polymerase initiation events, with noise added (bottom), or without (top). x-axis is time in minutes; y-axis is the intensity of transcription sites (expressed in number of RNA molecules). Right panels: positions of the transcription initiation events (vertical bars), for the original artificial data (black; bottom lines), the reconstructed data from the simulated short movies after the genetic algorithm (GA, red, middle lines), or the final reconstruction after both the GA and the local optimization (blue; top lines). x-axis is time in minutes. BE. Accuracy and robustness of the overall analysis pipeline. B The linear three-state promoter model used for Monte Carlo simulations. C Examples of artificial short movies (black lines), with various levels of noise added (red lines). Note that the experimentally measured noise level (resulting from the fitting deviations) corresponds to the 1x condition. x-axis is time in seconds; y-axis is the intensity of transcription sites expressed in number of RNA molecules. D Survival functions reconstructed from artificial short and long movies (red and green circles, respectively), and fitted to a sum of three exponentials (black line). The theoretical survival function obtained with the model parameters used for the simulation is shown for comparison (blue line). x-axis: time intervals between successive initiations events, in seconds and in log10 scale. y-axis: probability of Δt > x (log10 scale). E Accuracy of determining the model parameters. Graphs plot the parameters used to generate the artificial data (x-axis), against the parameter measured by the deconvolution and fitting procedure (y-axis). Vertical bars: confidence intervals estimated during the fitting procedure (see Methods). Three parameter sets were used, corresponding to the values obtained with the experimental data from the High Tat cells (circles), Low Tat cells (crosses), and No Tat cells (triangles).

Back to article page