Skip to main content
Fig. 4 | Genome Biology

Fig. 4

From: GRiNCH: simultaneous smoothing and detection of topological units of genome organization from sparse chromatin contact count matrices with matrix factorization

Fig. 4

Summary of benchmarking TAD-calling methods. Shown are different criteria of evaluation. A medal denotes whether the given TAD-calling method is among the top 3 methods for a particular criteria (gold/yellow: 1st place; silver/gray: 2nd place; bronze/brown: 3rd place). Validation: internal validation metrics for measuring the cohesiveness of predicted TADs. DBI: percentage of TADs with significant Davies-Bouldin index (Table S1a); DCC: percentage of TADs with significant Delta Contact Counts (Table S1b). Resolution: measuring stability of TADs to changing input data resolution (e.g., 10kb, 25kb, 50kb). Size: stability of median TAD size to Hi-C resolution (Table S1c); RI, MI: similarity of TADs from high- and low-resolution data, measured by Rand index (RI, Table S1d) and mutual information (MI, Table S1e). Depth: measuring stability of TADs to the depth and sparsity of input data. RI, MI: similarity of TADs from high-depth and low-depth data, measured by Rand index (RI, Table S1f) and mutual information (MI, Table S1g) Consistency: a group of methods yielding TADs with highest similarity, with gold for the pair of methods with highest similarity according to hierarchical clustering. Enrichment: measuring enrichment of regulatory signals. CTCF: fold enrichment of CTCF and cohesin elements in TAD boundaries (Table S1h); Histone: proportion of TADs with significant mean histone signal (Table S1i). Supplementary Tables S1a-i are available in Additional File 3

Back to article page