Extended Data Fig. 8. Annotation pipeline for feature pair presence and evaluation of sets with 10% of strongest contacts.

a, Strategy for detecting enrichment of feature pairs within sets of specific contacts. b, Pairwise differences of feature pair frequencies observed in the 10% sets of GAM-specific and Hi-C-specific contacts relative to their relative enrichment. Differences are colored according to the Gini impurity obtained by Random Forest predictions. c, Feature pairs with the highest discriminatory power in the 10% subsets relative to the abundance of the feature pair in the contact sets. Pairs with different order relative to the 5% results set italic.