SpectralTAD overview and hierarchy of TADs and boundaries. a) First two eigenvectors from decomposition of submatrices defined by sliding window across the diagonal of Hi-C matrix are projected on unit circle, and clustered based on largest gaps (defined by Euclidean distance) between eigenvectors. Silhouette score is used for automatic number of clusters (TADs) selection. b) TADs not enclosed within other TADs are defined as primary, while TADs contained within other TADs are defined as secondary, tertiary, etc. c) Boundaries are defined as the rightmost point of a given TAD. Boundaries belonging to a single TADs are defined as level 1, while boundaries shared by two, three, TADs are defined as level 2, 3, etc.