Table 2

Performance comparison based on a simulation dataset.

Number of mapped reads

Recall

Precision

Time(min)

Memory(MB)


BOAT

3,833,479

76.56%

99.41%

18

1,217

RMAP

2,957,658

58.89%

98.90%

840

2,371

SOAP

2,872,535

56.75%

97.19%

9

186**

MAQ

2,878,570

55.93%

93.53%

4

1,959

SeqMap*

2,187,611

43.57%

99.25%

33

12,500


5,000,000 simulated reads were mapped to an original two-million-bp mouse chrX region on a local Linux box with two Intel quad-core (E7310 @ 1.6 G Hz) CPUs and 64 G RAM. All programs were tuned to maximize their capability for tolerating no more than five mismatches (detailed running parameters for each tool were shown in Supplementary Table S2 of Additional File 2).

* We tried to run SeqMap with up to 5 mismatches, but failed with out-of-memory error. So only 3 mismatches with 1 indel were allowed when running SeqMap.

** As only a small part of the whole genome was used as reference sequence in this benchmark, the memory usage of SOAP is very low. However, when mapping to the whole human genome, at least 14 GB memory is required to run SOAP [5].

Zhao et al. BMC Genomics 2009 10(Suppl 3):S2   doi:10.1186/1471-2164-10-S3-S2

Open Data