Fig. 1

Workflow of deTOKI. a The flow chart of deTOKI. The genome is split into 8 MB overlapping sliding windows. As an example, b–f shows the key steps for a window in chr19. The data are from IMR90 cells [14]. In b, each square in blue, red, and green represents a sliding window. In c, NMF was performed on the contact matrix of a window, and bins were clustered based on the factor matrix. After 10 rounds of NMF, a consensus map was generated (d), in which the detected clustering change points are shown in sawtooth, and the blue curve in the bottom panel represents the clustering rate of each bin. e The blue curve represents the silhouette coefficient corresponding to each alternative value of “n_components”. f The yellow square highlighted the middle 4-Mb region for which clustering change points (yellow stars) were reported as TAD-like domain boundaries