Table 4.

Performance comparison of different scaffolders based on varying the sequencing depth on the MBARC-26 dataset.

Input size (millions of reads)	Metric	No. of scaffolds	Length at 1 Mbp (bp)	Length at 10 Mbp (bp)	Length at 50 Mbp (bp)	CPU time	Peak memory (GB)
2	OPERA-LG	29,831	988,539	90,227	719	44 s	2.2
	metaSPAdes	61,592	594,287	92,217	783	NA	8.2
	MetaCarvel	29,883	699,981	90,014	718	58 s	2.1
4	OPERA-LG	22,952	1,257,853	168,019	4393	2 min 16 s	2.8
	metaSPAdes	49,199	1,635,634	190,132	3823	NA	10.1
	MetaCarvel	23,003	1,257,853	168,390	4374	2 min 48 s	3
8	OPERA-LG	21,866	1,257,855	393,755	34,351	3 min 13 s	4
	metaSPAdes	39,460	1,635,634	190,132	31,823	NA	14.3
	MetaCarvel	23,003	1,223,449	423,739	32,331	3 min 47 s	3.8
10	OPERA-LG	21,413	1,257,855	402,996	50,874	8 min 01 s	5
	metaSPAdes	35,754	1,635,634	478,925	52,165	NA	22.5
	MetaCarvel	21,033	1,332,109	418,821	49,839	10 min 41 s	5.2
14	OPERA-LG	18,370	1,461,964	676,339	72,581	14 min 08 s	8.1
	metaSPAdes	29,298	1,635,789	668,856	78,337	NA	28.1
	MetaCarvel	18,281	1,463,318	686,311	73,522	13 min 14 s	7.4

The runtime for metaSPAdes is not mentioned (marked NA) since we cannot separate the assembly from the scaffolding steps. Maximum contig size is the same for OPERA-LG and MetaCarvel because the same input assembly was used as input to them