Figure 5. Hub sharing was limited to adjacent time scales.

(A) We classified each neuron as a hub, non-hub, or unconnected neuron at each time scale. A neuron was considered to be a shared hub or shared non-hub for two time scales if its status as a hub or non-hub was consistent across those time scales. Hubs were defined using a degree threshold set by the likelihood to have a given number of connections in a random network (0.05 in this illustrative diagram and 10⁻⁴ in the full analysis). (B) We calculated the amount of hub and non-hub sharing (see Materials and Methods) for each pair of time scales and grouped the results into neighboring (4 or less) and distant (greater than 4) time scales. We found that hubs were only shared at a significant level for neighboring time scales, while non-hubs were broadly shared across all time scales (multiple comparisons correct Mann-Whitney Test (1, 2, and 3 dots: p<0.05, 0.01, and 0.001 respectively), error bar: standard error of the mean). For each data set, we subtracted the mean sharing values for 500 trials with neuron identities randomized and neuron hub, non-hub, or unconnected status held constant. This null model approximates the amount of sharing expected based only on the number of hubs, non-hubs, and unconnected neurons in the data set, as well as the effect of ignoring the multiplex properties of the networks and considering the time scales to be truly independent networks. We also calculated the mean sharing value of (C) hubs and (E) non-hubs across each pair of time scales for cortical and hippocampal networks. In (B), neighboring time scale pairs are up and to the left of the white line, while distant time scale pairs are down and to the right of the white line. (D and F) Finally, we calculated the multiple comparisons corrected Mann-Whitney Test p-values between sharing results from data and sharing results from the null model.