Fig. 5

Comparison of clustering accuracy for 6 Hi-C scaffolding tools. LACHESIS and ALLHiC were run with chromosome number as input, and the other Hi-C scaffolding tools do not need the chromosome number. For human (n = 23), rice (n = 12), Arabidopisis (n = 5), the reference genomes of human CHM13 v1.1, rice (Nipponbare) ASM386523v1, Arabidopsis thaliana (Columbia) TAIR10.1 were used for comparison with the results of four Hi-C scaffolders. For great burdock (n = 18) and water spinach (n = 15), the genomes of closely related species Cynara cardunculus (NCBI RefSeq GCF_001531365.1) and Ipomoea triloba (NCBI RefSeq GCF_003576645.1) were used for comparison. For fairness, only larger contigs with size > 1 Mb were used for all the 6 Hi-C scaffolding tools, and the percent of used contigs were shown in Table 2. Just one round of EndHiC was applied to all the 3 simulated datasets (human, rice, Arabidopsis) and two real datasets (great burdock, water spinach), and two rounds of EndHiC were applied to the human real dataset. Complete: A Hi-C scaffold includes most fragments only from one chromosome, i.e. the total length of shared contigs between a Hi-C scaffold (cluster) and the reference chromosome accounts for over 85% of both the scaffold (cluster) and the chromosome. Segment: A chromosome is fragmented into two or more large scaffold, and each segment scaffold accounts for 5–85% of the corresponding chromosome. Mis-joined: A Hi-C scaffold includes large fragments from two or more chromosomes. In overall, the clustering accuracy for these large contigs is: LACHESIS < ALLHiC < 3D-DNA < Pin_hic < YaHS < EndHiC