Skip to main content

Table 5 FDRs of evaluated methods

From: UMI-count modeling and differential expression analysis for single-cell RNA sequencing

  

1a

 

0.8–0.9b

0.5–0.6c

Method

FDR

False (n)

DE (n)

FDR

False (n)

DE (n)

FDR

False (n)

DE (n)

Monocle2

0.069

5.9

83.9

0.089

7

79.1

0.276

22.1

79

SCDE

0.299

2.6

8.3

0.34

3.7

9.5

0.848

123.5

145.2

MAST

0.001

0

29.5

0.003

0.1

28.2

0.193

3.4

19.5

ROTS

0.045

4.4

97.5

0.497

71.6

145.9

0.835

272.4

323.9

Seurat_ttest

0.244

31.5

128.3

0.441

69.6

156.1

0.927

653.6

704.5

Seurat_bimod

0.154

17.6

112.5

0.655

172

258.7

0.928

924.5

996.3

Seurat_tobit

0.248

32.2

129

0.45

72

158.5

0.873

351.7

402.8

Seurat_poisson

0.208

25.7

122.6

0.188

20.1

106.1

0.573

67.4

116.2

Seurat_negbinom

0.197

23.9

120.7

0.164

16.9

102.4

0.5

47.7

93.8

NBID_scran

0.038

3.4

86.9

0.035

2.8

80.1

0.039

2.5

62.4

NBID

0.033

2.8

85.7

0.032

2.7

81.4

0.03

1.8

58.9

  1. aNo sub-sampling
  2. bThe sub-sampling ratio in Group 2 was 0.8–0.9
  3. cThe sub-sampling ratio in Group 2 was 0.5–0.6.
  4. Bold values indicate FDR > 0.05. Bold and underlined values indicate FDR > 0.1. The nominal FDR was 0.05. Simulation based on the Memory T-cell data [8], 500 cells in each group, results are averaged over 96 replicates (see Additional file 2: Tables S5–S9 for results for other simulation scenarios). NBID_scran used the size factor computed by scran as the offset instead of the total UMI counts
Back to article page

Genome Biology

ISSN: 1474-760X

Contact us