Table 2 A comparison between the results of six methods: proposed, RaceID, SC3, Seurat, SINCERA, and SNN-Cliq.

From: Identification of cell types from single cell data using stable clustering

Dataset

#cell types

Proposed

RaceID

SC3

SINCERA

SNN-Cliq

Seurat

K (mean ± sd)

AMI (mean ± sd)

K (mean ± sd)

AMI (mean ± sd)

K (mean ± sd)

AMI (mean ± sd)

K

AMI

K

AMI

K

AMI

Biase

3

3 ± 0

0.92 ± 0.02

3.14 ± 0.6

0.85 ± 0.23

3 ± 0

0.92 ± 0

6

0.64

6

0.62

4

0.74

Deng

10

10 ± 0

0.73 ± 0.01

1 ± 0

0 ± 0

9 ± 0

0.81 ± 0.006

3

0.48

17

0.6

6

0.59

Goolam

5

3 ± 0

0.73 ± 0.04

1 ± 0

0 ± 0

6 ± 0

0.69 ± 0

13

0.4

17

0.42

3

0.11

Klein

4

6 ± 0

0.67 ± 0.06

2.98 ± 0.14

0.51 ± 0.05

19 ± 0

0.53 ± 0.006

43

0.52

265

0.21

3

0.06

Patel

5

5 ± 0

0.86 ± 0.01

7.44 ± 1.88

0.66 ± 0.1

17 ± 0

0.93 ± 0

10

0.73

26

0.31

5

0.68

Pollen

11

8 ± 0

0.72 ± 0.01

8.36 ± 2.27

0.68 ± 0

10 ± 0

0.53 ± 0.01

10

0.91

22

0.74

8

0.87

Treutlein

5

3 ± 0

0.54 ± 0.03

1 ± 0

0 ± 0

3 ± 0

0.62 ± 0

7

0.46

5

0.51

1

0

Yan

8

5 ± 0

0.78 ± 0.01

5.5 ± 2.34

0.61 ± 0.17

4 ± 0

0.72 ± 0

8

0.72

13

0.76

3

0.58

sim3

3

3 ± 0

1 ± 0

1 ± 0

0 ± 0

3 ± 0

1 ± 0

120

0.23

147

0.21

3

1

sim4

4

4 ± 0

0.99 ± 0.007

1 ± 0

0 ± 0

4 ± 0

0.99 ± 0.001

464

0.21

437

0.2

3

0.66

sim6

6

7.9 ± 0.3

0.64 ± 0.02

1 ± 0

0 ± 0

3 ± 0

0.51 ± 0.004

68

0.42

143

0.3

6

1

sim8

8

9.34 ± 0.47

0.85 ± 0.01

1 ± 0

0 ± 0

4 ± 0

0.56 ± 0.007

68

0.51

290

0.31

8

1

sim_Tung

8

8 ± 0

0.51 ± 0.008

1 ± 0

0 ± 0

8 ± 0

0.006 ± 0

17

0.04

77

0.13

8

0

  1. The adjusted mutual information (AMI)48,49, is used to evaluate the performance of each clustering method. The proposed method, RaceID, and SC3 are performed 50, 50, and 5 times on each dataset, respectively. The average AMIs across different runs are computed for the proposed method, SC3, and RaceID. Since SNN-Cliq, SINCERA and SEURAT are deterministic, they are performed only once.