Fig. 2: Network hubs are robust across parcellation schemes, ages and individuals.

A Degree distributions (top row) are shown at two alternative resolutions (µBrain₆₀ and µBrain₁₂₀) with subparcels of ~60 and 120 vertices in area, respectively, with a threshold of 15%. Corresponding rich-club (RC) coefficient curves (middle row) are shown for each as in Fig. 1. Normalised rich-club coefficient (red line) values > 1 indicate greater rich-club organisation than expected by chance. The locations of hub nodes (bottom row) in the left and right hemispheres (coloured regions) are shown overlaid on each parcellation resolution. B Correlation between nodal degree profiles in average networks generated from subsamples of increasing size and the whole-group network. The red line indicates the average across the 100 random subsamples, while the shaded red area shows the standard deviation. C Correlations between nodal degree of group-averaged networks (µBrain₉₀ thresholded at 15%) created using all individuals scanned within a particular gestational week (ranging from the 37th week to the 44th week) and the whole-group average network. D Top, for each individual neonate’s structural network, we identified hub nodes (90th percentile for degree). We then calculated the proportion of times a node was identified as a hub across individuals (hubness across individuals). Bottom, individual hubness plotted against the group-averaged degree. Points in red were identified as a hub in the group-averaged consensus network (90th percentile for degree).