Fig. 5

Differential regulatory units across mouse neural differentiation. a Interaction matrices (log O/E) of three Hi-C experiments of mouse embryonic stem cells (ES), neural progenitor cells (NPC), and cortical neurons (CN). A differential regulatory unit is indicated with dark blue rectangles, showing the interaction of two dynamic active enhancer regions (light blue) and the correlated gene Inhbb. b Differentially active enhancers were filtered for regions that are only active in ES (I), only active in NPC (II), and only active in CN (III). For these regions, normalized (log O/E) chromatin interaction counts that overlap the predicted differential regulatory units were re-scaled to [0,1], such that the highest interaction count for each region is 1. The results of all regulatory units (yellow) are compared to a subset where just target genes with the highest correlation were taken into account (orange). Additionally, for each differential enhancer in this subset, the nearest gene was also taken as an alternative target (blue). All three methods reflect the expected dynamic behavior, meaning that the scaled interaction counts are close to 1 for the respective highlighted condition but lower in the other two conditions. It also becomes apparent that choosing the closest gene as a target might not always be the best choice as can be seen for the regions I and II