Figure 1

Linearity of the relation phase/HL in the clustered YMC. In (a) the time series (on the x axis: time in min.) of the periodic genes is clustered according to a nonnormalized correlation distance function (see Table 1 for details on the clusters). The clusters are then sorted (from left to right from top to bottom) according to the average HL. Moving along the clusters, a change in the phase and in the width of the pulses is clearly visible, thus suggesting a direct relationship between HL and phase/width of the pulses. This is made explicit in (b), where the average HL is plotted against the average phase for each cluster, and in (c) where the average HL is shown against the average pulse width. In the scatter plot of HL versus phase (d), the color indicates the cluster number (see colorbar on the right). As can be noticed, along the HL axis the standard deviation of a cluster grows with the mean, see Table 1 for exact values, and the cloud of points looks like a cone (the cone delimited by the two red lines contains 95% of the periodic genes). Still the increase of the phase with the HL is clearly visible. In the least-squares linear fit in (b) (green) half of the L₂ norm of the residues is due to Cluster 9 (cytoplasmic ribosomes, see text). The p. value for both linear regressions is < 10^-5. Further details on these regressions are provided in Additional file 1. It is worth remarking that the direct proportionality phase/HL is robust with respect to the number of clusters chosen.