Skip to main content

Table 1 Number of DEGs Identified for Case Studies

From: Filtering for increased power for microarray data analysis

   

Filtering Method

Case Study

Preprocessing Method

FDR Method

None

Detection Call

Variance (log2)

Variance (original)

Signal

Wheat Comp 1

RMA

None

15511

12524

12601

12932

12459

  

BH

8546

9355

9464

9725

9163

  

q-value

10333

12237

12369

13125

12233

Wheat Comp 1

MAS5

None

12547

10520

3497

10728

10539

  

BH

5869

6896

1096

7088

6880

  

q-value

6687

8968

1137

9305

8989

Wheat Comp 1

PLIER

None

17224

13127

8151

13992

13305

  

BH

9713

9811

5436

10714

9913

  

q-value

12167

13366

6010

15541

14242

Wheat Comp 2

RMA

None

10309

7454

7751

7776

7391

  

BH

643

1593

1733

1704

1329

  

q-value

1416

3390

3874

3791

3175

Wheat Comp 2

MAS5

None

8149

6187

2620

6353

6201

  

BH

162

569

42

585

511

  

q-value

328

1337

53

1485

1330

Wheat Comp 2

PLIER

None

10928

7473

5690

8150

7539

  

BH

1048

1654

1219

2109

1633

  

q-value

2067

3386

1736

4402

3366

Diabetes

RMA

None

3469

3043

2710

3018

2991

  

BH

644

781

710

807

783

  

q-value

728

892

778

904

884

Diabetes

MAS5

None

3097

2599

1555

2597

2595

  

BH

412

481

239

484

478

  

q-value

449

520

241

523

517

Diabetes

PLIER

None

3266

2895

1792

2912

2921

  

BH

541

646

388

682

665

  

q-value

615

765

440

793

759

Smoking

RMA

None

3414

2511

2155

2196

2146

  

BH

126

155

143

135

107

  

q-value

135

184

177

158

121

Smoking

MAS5

None

3684

2522

1843

2510

2506

  

BH

112

145

109

146

137

  

q-value

130

166

116

164

155

Smoking

PLIER

None

3097

1832

1824

1704

1643

  

BH

64

80

78

77

70

  

q-value

70

91

85

81

72

  1. Table of number of probe sets identified as differentially expressed for each of the case studies. For each case study, we considered three preprocessing methods (RMA, MAS5 and PLIER), two FDR methods (Benjamini-Hochberg and q-value) and three filtering methods (MAS detection call, variance (on both the log2 and original scales) and signal). The stated significance level was 0.05 for all methods.