Fig. 4

Visualization of the networks from our study. Illustration of SIGN N (signed Pearson correlation with top N interactions), ABS N (absolute Pearson correlation with top N interactions), MI N (mutual information), and ABS-MI-10 N (intersection of absolute Pearson correlation and mutual information with top 10 N interactions) networks. Intersection of absolute Pearson correlation and mutual information with top 10 N interactions is not shown, as it resembles ABS-MI-10 N. In our study, for prediction purposes, we study the networks’ largest connected components. In blue we show the subset of all genes from the given network that are among the 1,661 rhythmic genes determined using JTK_CYCLE statistical analyses of the expression data. In yellow we show the subset of all genes from the given network that are among the novel network-based predicted rhythmic genes, i.e., network-based predicted rhythmic genes that could not be identified using the JTK_CYCLE statistical analyses. Note that the differences between some of these networks should not be surprising, since the networks were constructed using different network inference approaches. Also, note that the network visualizations are only intended for illustration. One should not rely on visualizations to determine how meaningful the networks are. For example, what appears to be a group (cluster) of rhythmic genes in the given figure/network might not be reported as a statistically significantly meaningful cluster by network analysis. Or, what appears to be a single cluster in the figure might be broken down and reported as multiple clusters by network analysis. Network analysis (rather than visualization) is a systematic and mathematically/computationally non-ambiguously precise way of interpreting the network data