Skip to main content

Table 2 Performance measures of KMA, SRST2, MGmapper, BWA-MEM, Bowtie2, Minimap2 and Salmon, for predicting genes directly from raw reads. Thresholds for predicting a gene has been set to: 90% coverage, 90% identity and a minimum depth of 5. A minimum mapping quality of 10 was used for methods relying on post processing with SAMtools and BEDTools, as this gave the best performance across the tested thresholds

From: Rapid and precise alignment of raw reads against redundant databases with KMA

Mapping method

Post- processing method

Avg. mapping CPU time

Avg. post- processing CPU time

Peak memory

MCC

Predicting antimicrobial resistance

 KMA

NA

00:00:24.6

NA

42.3 MB

1.000

 SRST2

NA

00:10:21.3

NA

165.0 MB

1.000

 MGmappera

NA

00:13:14.2

NA

101.4 MB

0.288

 BWA-MEM

SAMtools / BEDTools

00:07:35.5

00:00:06.1

113.0 MB

0.000

 BWA-MEMb

Salmon

00:07:34.5

00:00:12.2

694.9 MB

0.828

 Bowtie2

SAMtools / BEDTools

00:02:35.5

00:00:06.7

33.7 MB

0.000

 Bowtie2b

Salmon

00:03:16.4

00:02:24.5

935.8 MB

0.623

 Minimap2

SAMtools / BEDTools

00:02:18.6

00:00:06.0

517.3 MB

0.000

Mapping towards cgMLST alleles

 KMA

NA

00:07:02.1

NA

8.3 GB

0.998

 SRST2

NA

> 99:99:99.9

NA

NA

NA

 MGmappera

NA

01:23:21.5

NA

8.7 GB

0.062

 BWA-MEM

SAMtools / BEDTools

02:14:50.8

00:14:06.7

8.9 GB

0.021

 BWA-MEMb

Salmon

03:22:45.4

04:41:09.6

104.2 GBP

0.530

 Bowtie2

SAMtools / BEDTools

01:50:56.8

00:15:23.5

4.1 GB

0.035

 Bowtie2b

Salmon

> 99:99:99.9

NA

NA

NA

 Minimap2

SAMtools / BEDTools

01:20:56.2

00:13:11.7

33.6 GB

0.035

  1. a MGmapper was executed on the forward reads only, as paired end mode crashed
  2. b Report all alignments
  3. P The post processing method was responsible for the peak memory consumption