Table 1 Performance of KSNP and the four state-of-the-art haplotyping tools on PacBio CLR and HiFi datasets

From: KSNP: a fast de Bruijn graph-based haplotyping tool approaching data-in time cost

Dataset22,25

Toola

SE (%)

HE (%)

Hap N50 (kb)

Recall (%)

CPU time (s)

Wall time (s)

RAM (MB)

HG001 CLR 50×

Longshot

0.63

1.14

239

91.56

13,485

14,186

2970

WhatsHap

0.67

1.57

241

91.61

20,714

22,050

1843

HapCUT2

0.67

1.84

251

91.66

6100

6478

969

Margin

0.63

0.89

171

90.30

146,551

19,283

2867

KSNPa

0.68

1.82

246

91.59

1881

2004

510

HG002 CLR 50×

Longshot

1.22

1.53

315

89.86

13,537

14,267

2662

WhatsHap

1.25

2.03

317

89.91

19,141

19,963

1536

HapCUT2

1.25

2.25

324

89.95

6042

6453

802

Margin

1.21

1.38

244

89.17

128,479

16,471

2458

KSNP

1.25

2.12

321

89.91

1607

1740

476

HG002

HiFi 50×

Longshot

1.26

1.21

416

93.67

11,285

11,982

2458

WhatsHap

1.27

1.48

417

93.68

13,254

13,880

1229

HapCUT2

1.27

1.41

417

93.68

5125

5483

685

Margin

1.26

1.17

370

93.42

112,980

15,064

2253

KSNP

1.27

1.46

422

93.68

1106

1270

462

HG005 CLR 50×

Longshot

1.45

2.77

490

92.75

17,968

18,865

2765

WhatsHap

1.49

4.01

507

92.80

22,968

24,093

1434

HapCUT2

1.49

4.43

528

92.84

8202

8659

946

Margin

1.44

2.40

315

91.23

145,455

18,648

2253

KSNP

1.49

3.98

514

92.80

2088

2268

513

HG01109 CLR 40×

Longshot

0.04

4.92

5080

89.61

20,794

21,930

2662

WhatsHap

0.06

5.42

5323

89.66

22,919

23,917

1331

HapCUT2

0.07

6.11

5717

89.67

9615

10,184

849

Margin

0.03

2.46

3804

89.47

125,506

16,513

2355

KSNP

0.07

5.94

5333

89.66

1304

1449

512

A.thaliana CLR 45×

Longshot

0.01

2.59

4001

85.97

3570

3586

911

WhatsHap

0.01

2.12

4001

85.98

2610

2655

615

HapCUT2

0.01

2.43

4001

85.99

1753

1759

392

Margin

0.01

2.07

1228

85.37

9619

1202

2150

KSNP

0.01

2.56

4001

85.98

100

100

152

  1. SE switch error rate, HE hamming error rate, Wall time wall clock time, RAM peak RAM.
  2. aLongshot, WhatsHap, HapCUT2, and KSNP were performed with one thread in the experiments, while Margin utilized eight threads. The k value in KSNP was set to two by default.