Figure 2. Structure of noisiness shows partitioning between day and night.

1. Number of genes identified as being highly variable for each time‐point. These genes are separated between those that are also selected as highly variable in at least one other time‐point (dark blue) and those highly variable in only one time‐point (light blue). The top bar indicates time‐points harvested during the day (orange), just before dusk (red), during the night (black) and just before dawn (blue).
2. Distribution of the number of time‐points at which genes are identified as highly variable (blue) or lowly variable (green). The distribution of average (black) and 95% confidence interval (dotted grey) for the thousand random sets are also represented and are so close that they are superimposed and cannot be differentiated in the figure.
3. Heatmap of the percentage of HVGs shared between time‐points. Red indicates a high percentage of HVGs in common between two time‐points. The top and side bars indicate time‐points harvested during the day (orange), just before dusk (red), during the night (black) and just before dawn (blue).
4. Hierarchical clustering of HVGs based on the log₂(CV²/trend) at each time‐point. The result is represented as a heatmap where yellow indicates a high log₂(CV²/trend). The genes were separated into four clusters, indicated by the side coloured bar. The top bar indicates time‐points harvested during the day (orange), just before dusk (red), during the night (black) and just before dawn (blue). See Appendix Fig S3G for heatmaps of the log₂(CV²/trend) with the same colour cut‐offs for HVGs, LVGs and random genes.
5. Heatmap of the mean normalised expression level for the genes in Fig 2D, keeping the same clustering organisation. The top bar indicates time‐points harvested during the day (orange), just before dusk (red), during the night (black) and just before dawn (blue).