Table 4.

Overall evaluation and comparison of multiple aligners.

Aligners	Computational speed			Memory usage			Accuracy
	Speed with single thread	Speed with multithread	Key factor impacting speed (genome size or read count)	Overall evaluation	Key factor impacting memory (Genome size or read count)	Memory usage with multithread	Sensitivity	Precision	% of multimapped	%Corrected Multi-Mapped
Bowtie1	Fast	↑	Genome size	Low	Genome size	≡	High	—	—
BWA	Fast	↑	Both	Low	Genome size	≡
BOAT	Slow	↑↑	Genome size	Low	Read count	↑↑	High	—	—	Low
GASSST	—	↑	Genome size	High★★	Genome size	≡	Low	High	—
Gnumap	Slow	↓	Genome size	High★★	Genome size	≡
GenomeMapper	Slow	≡	Genome size	Low▲	Genome size	≡	High	—	—
mrFAST	Slow	×	Genome size	High★★	Read count	×	High	—	—
mrsFAST	—	×	Genome size	Low	Read count	×	High	—	—
MAQ	—	×	Genome size	High★★	Read count	×
NovoAlign#	—	/	Read count	Low▲	Genome size	/	High	High	Low	Low
PASS	—	↑	Genome size	Low▲	Genome size	↑	High	High	Low	Low
PerM※	Fast		Genome size	Low▲	Genome size	/	Ind: low	—	Low
RazerS	Slow	×	Genome size	High★★	Read count	×	High	—	—
RMAP	—	×	Genome size	High★	Genome size	×	Mis: low	High	Low
SeqMap	—	×	Genome size	High★★★	Read count	×	High	—	—
SOAPv2	Fast	↑	Genome size	Low	Genome size	≡	High	High	Low
SHRiMAP2	Slow	↑	Genome size	High★★	Genome size	↑	High	Low	High
Segemehl	—	↑	Both	High★★★	Genome size	≡	High	—

PerM^※ could adjust the threads automatically during running process.

Novoalign^# could support multithread only for commercial version.

For computational speed, we defined the aligners which are extremely faster than others as fast, while we defined the ones which are extremely slower as slow.

For memory usage, we evaluated the aligners as follow: among the s even datasets, the maximum memory usage ≤4 G, low; the maximum memory usage ≥32 G, high^★★★.

Low^▲ represents that the maximum memory usage will have an extreme increase with H. sapiens datasets (≥4 G).

×: without multithread function.

— represents medium level remark.

≡ means there is no obvious change.