Table 2:

Benchmarking leading solutions against SeQuiLa-cov on WES/WGS data in execution time of block and window calculations

Data	Operation type	Cores	samtools	bedtools	sambamba	mosdepth	SeQuiLa-cov
WGS	Blocks	1	2h 14m 58s1	10h 41m 27s	2h 44m 0s	1h 46m 27s	1h 47m 5s
		5			2h 47m 53s	36m 13s	26m 59s
		10			2h 50m 47s	34m 34s	13m 54s
	Fixed-length windows	1			1h 46m 50s	1h 22m 49s	1h 24m 8s
		5			1h 41m 23s	20m 3s	18m 43s
		10			1h 50m 35s	17m 49s	9m 14s
WES	Blocks	1	12m 26s1	23m 25s	25m 42s	6m 43s	6m 54s
		5			25m 46s	2m 25s	1m 47s
		10			25m 49s	2m 20s	1m 4s
	Fixed-length windows	1			14m 36s	6m 11s	6m 29s
		5			14m 54s	2m 8s	1m 42s
		10			14m 40s	2m 14s	1m 1s

Both samtools and bedtools calculate coverage using only a single thread; however, their results differ significantly, with samtools being approximately twice as fast. Sambamba positions itself as a multithreaded solution, although our tests revealed that its execution time is nearly constant, regardless of the number of CPU cores used, and even twice as slow as samtools. Mosdepth achieved speedup against samtools in blocks coverage and against sambamba in windows coverage calculations; however, its scalability reaches its limit at 5 CPU cores. Finally, SeQuiLa-cov achieves performance nearly identical to that of mosdepth for the single core, but the execution time decreases substantially for greater number of available computing resources, which makes this solution the fastest when run on multiple cores and nodes.

¹Per-base results are treated as block output. Samtools lacks the functionality of block coverage calculations; however, we included this tool in our benchmark for completeness, treating its per-base results as block outcome assuming that both result types require nearly the same resources.