Table 3 Performance of algorithms in the sample-classification challenges on the validation cohorta

From: Critical assessment of automated flow cytometry data analysis techniques

 

Recall

Precision

Accuracy

F-measure

Recall

Precision

Accuracy

F-measure

Recall

Precision

Accuracy

F-measure

 

Challenge 1: HEUvsUE

Challenge 2: AML

Challenge 3: HVTN

FlowCAP

2DhistsSVMb

0.50

0.091

0.50

0.15

0.00

0.95

0.99

0.97

    

EMMIXCYTOM

    

0.95

0.95

0.99

0.95

    

flowBin

0.012

0.00

0.45

0.00

0.10

0.30

0.92

0.46

    

flowCore-flowStats

0.56

0.455

0.55

0.50

    

1.00

1.00

1.00

1.00

flowPeakssvm

    

1.00

1.00

1.00

1.00

    

flowType

0.58

0.636

0.59

0.61

0.95

0.95

0.99

0.95

0.88

0.71

0.81

0.79

flowType-FeaLect

0.55

0.545

0.55

0.55

1.00

1.00

1.00

1.00

1.00

1.00

1.00

1.00

Kmeanssvm

        

1.00

1.00

1.00

1.00

PBSC

0.33

0.273

0.36

0.30

0.75

0.75

0.94

0.75

0.95

0.95

0.95

0.95

PRAMS

        

1.00

1.00

1.00

1.00

Pram Spheres

0.36

0.364

0.36

0.36

    

0.90

0.90

0.90

0.90

Random Spheres

    

0.95

0.95

0.99

0.95

    

SPADE

    

1.00

1.00

1.00

1.00

1.00

1.00

1.00

1.00

SWIFT

0.67

0.545

0.64

0.60

    

1.00

1.00

1.00

1.00

DREAM

            

admire-lvq

    

1.00

1.00

1.00

1.00

    

bcb

    

1.00

1.00

1.00

1.00

    

biolobe

    

1.00

1.00

1.00

1.00

    

cihc

    

1.00

0.95

0.99

0.97

    

daltons

    

1.00

1.00

1.00

1.00

    

DREAM–A

    

0.95

0.95

0.99

0.95

    

DREAM–B

    

1.00

0.85

0.98

0.92

    

DREAM–C

    

1.00

0.85

0.98

0.92

    

DREAM–D

    

0.95

0.95

0.99

0.95

    

fivebyfive

    

0.95

1.00

0.99

0.98

    

jkjg

    

1.00

1.00

1.00

1.00

    

SPCA+GLM

    

0.89

0.85

0.97

0.87

    

team21

    

1.00

1.00

1.00

1.00

    

uqs

    

1.00

0.95

0.99

0.97

    
  1. aNot all algorithms were applied in all challenges. Particularly, a large number of algorithms participated through the DREAM project that included only the AML data set. Data sets: HEUvsUE, HIV-exposed–uninfected versus unexposed; AML, acute myeloid leukemia; HVTN, identification of antigen stimulation groups of post–HIV vaccine T cells.
  2. bContact information of the participating teams can be found in Supplementary Table 3.