Skip to main content

Table 2 Performance of detection methods on datasets HG002 CLR and HG002 CCS

From: cnnLSV: detecting structural variants by encoding long-read alignment information and convolutional neural network

Coverage

Metric (%)

Sniffles

PBSV

SVIM

cuteSV

cnnLSV

HG002 CLR

69 \(\times\)

Pre

94.364

94.825

94.981

95.233

95.976

Rec

89.265

87.574

91.899

91.547

93.403

F1

91.744

91.055

93.415

93.354

94.672

40 \(\times\)

Pre

93.127

94.782

95.343

95.253

95.512

Rec

90.177

86.423

88.352

90.966

91.754

F1

91.628

90.410

91.715

93.060

93.595

30 \(\times\)

Pre

91.710

95.001

96.277

94.941

95.143

Rec

88.684

84.016

82.211

88.528

89.410

F1

90.171

89.172

88.690

91.623

92.187

20 \(\times\)

Pre

92.845

95.427

85.530

96.193

92.835

Rec

77.316

76.921

90.374

77.471

88.248

F1

84.372

85.181

87.886

85.823

90.483

10 \(\times\)

Pre

90.322

96.888

91.837

96.365

92.161

Rec

57.629

49.590

73.685

57.380

73.654

F1

70.363

65.603

81.766

71.930

81.875

HG002 CCS

28 \(\times\)

Pre

94.059

93.622

93.312

94.964

94.858

Rec

93.777

84.172

92.770

93.393

93.673

F1

93.918

88.646

93.040

94.172

94.261

10 \(\times\)

Pre

94.092

95.468

92.192

94.937

93.748

Rec

88.373

75.532

90.343

87.014

90.872

F1

91.143

84.338

91.258

90.803

92.288

  1. The values in bold represent the best results