EndHiC: assemble large contigs into chromosome-level scaffolds using the Hi-C links from contig ends

BMC Bioinformatics

Table 3 Performance comparison for EndHiC with other 5 Hi-C scaffolding tools

Hi-C scaffolder	Human			Rice			Arabidopsis
Hi-C scaffolder	Elapsed time (m)	CPU time (m)	Peak memory (Gb)	Elapsed time (m)	CPU time (m)	Peak memory (Gb)	Elapsed time (m)	CPU time (m)	Peak memory (Gb)
LACHESIS	5.6	5.6	2.7	9.9	9.9	0.4	1.5	1.5	0.30
ALLHiC	22.3	43.2	38.2	17.1	40.7	16.8	3.2	7.0	2.80
3D-DNA	1,314.0	2,100.0	51.6	354.0	816.0	51.6	102.0	534.0	28.70
Pin_hic	171.7	150.3	5.6	24.6	22.9	0.6	15.8	14.2	0.60
YaHS	29.7	29.4	7.0	5.0	3.7	0.4	1.6	1.6	0.14
EndHiC	0.5	3.6	0.2	0.2	1.1	0.1	0.1	0.6	0.02

All the statistics here do not include the preprocessing steps (reads mapping and filtering), so it is fair to compare the major Hi-C scaffolding algorithms among the 6 tools. The simulated contig data of three model species shown in Table 2 were used to test EndHiC, LACHESIS, ALLHiC, 3D-DNA, Pin_hic and YaHS. For all these simulated datasets, just one round of EndHiC was applied. Note that LACHESIS only finished the cluster step and then exited, it did not perform order and orientation here. The reason might be that LACHESIS grouped all the simulated large contigs into a single cluster and produced invalid intermediate files for later steps. Based on previous experience, the speed and memory consumption of LACHESIS is comparable to that of ALLHiC

ISSN: 1471-2105