Skip to main content

Table 5 Comparison between this study and Tox21 Data Challenge winners in terms of the classification performance metrics AUROC and balanced accuracy

From: Structure–activity relationship-based chemical classification of highly imbalanced Tox21 datasets

Assay ID

AUROC

Balanced accuracy (BA)

Best classifier / challenge winner

Best classifier (this study)

Dmlab [58]

Microsomes [59]

Challenge winner

Best classifier (this study)

Dmlab [58]

Microsomes [59]

Challenge winner

Value

Name

  

Value

Name

  

AUROC

BA

NR-AR

0.823

RF

0.830

N/A

0.828

0.644

SMN

0.610

N/A

0.736

0.994

0.875

NR-AR-LBD

0.913

RUS

0.820

N/A

0.879

0.612

RUS

0.490

N/A

0.650

1.039

0.942

NR-AhR

0.920

SMN

0.780

0.901

0.928

0.823

SMN

0.560

0.698

0.853

0.991

0.965

NR-Aromatase

0.850

SMN

0.840

N/A

0.838

0.727

SMN

0.560

N/A

0.737

1.014

0.986

NR-ER

0.863

SMN

0.770

0.783

0.810

0.792

SMN

0.660

0.621

0.749

1.065

1.057

NR-ER-LBD

0.823

SMN

0.770

0.827

0.827

0.697

RUS

0.590

0.550

0.715

0.995

0.975

NR-PPAR-γ

0.794

RUS

0.830

0.718

0.861

0.745

RUS

0.550

N/A

0.785

0.922

0.949

SR-ARE

0.891

SMN

0.770

0.804

0.840

0.855

SMN

0.520

0.605

0.729

1.061

1.173

SR-ATAD5

0.815

SMO

0.800

0.812

0.828

0.713

RUS

0.610

0.539

0.741

0.984

0.962

SR-HSE

0.848

SMN

0.860

N/A

0.865

0.667

RUS

0.560

N/A

0.799

0.980

0.835

SR-MMP

0.930

RUS

0.950

N/A

0.950

0.852

RUS

0.69

N/A

0.904

0.978

0.942

SR-p53

0.879

SMN

0.880

0.826

0.880

0.778

RUS

0.58

0.523

0.765

0.998

1.017

Average

0.862

 

0.830

0.810

0.861

0.742

 

0.58

0.589

0.764

1.002

0.973

  1. The values in italics are the highest among all the classifiers (both this study and Tox21 Data Challenge) whereas the values in bold font are the best among the Tox21 Data Challenge participating teams [34]