Systematic evaluation of 16 TAD-calling methods based on the boundary voting strategy. (A) Proportion of boundaries with different levels of boundary score for all 16 methods. These methods are sorted in ascending order by the proportion of boundaries belonging to the first level (ranging from one to two). (B) Profiles of three topological indicators (insulation score [IS], directionality index [DI], contrast index [CI]) and profiles of three structural proteins (CTCF, RAD21, SMC3) within 2-Mb regions centered on boundaries with different boundary score levels or randomly selected regions. (C) Number of boundaries with different boundary scores captured or missed by each method. These methods are sorted in ascending order by the number of captured boundaries. (D) Hi-C contact maps around the unreliable TADs with low boundary scores for three methods, including OnTAD, GMAP, and CaTCH. The unreliable TADs are indicated by white frames, and the low-scoring boundaries are indicated by dashed lines. (E) Hi-C contact maps around the high-scoring boundaries that are missed by a certain method. The examples for three methods including OnTAD, GMAP, and CaTCH are shown. The missed boundaries are indicated by red arrows, and the corresponding boundary scores are indicated by red triangles.
