Coeliac vs. refractory 100 features

Sparse logistic regression (ROC)

AUC = 0.8 (0.6–1)

sensitivity = 0.62 (0.32–0.86)

specificity = 1 (0.59–1)

PPV = 1

NPV = 0.58

p-Value = 0.013

Random forest (ROC)

AUC = 0.81 (0.62–1)

sensitivity = 0.62 (0.32–0.86)

specificity = 1 (0.59–1)

PPV = 1

NPV = 0.58

p-Value = 0.011

Gaussian process (ROC)

AUC = 0.92 (0.81–1)

sensitivity = 0.77 (0.46–0.95)

specificity = 1 (0.59–1)

PPV = 1

NPV = 0.7

p-Value = 0.000

Support vector machine (ROC)

AUC = 0.91 (0.79–1)

sensitivity = 0.85 (0.55–0.98)

specificity = 0.86 (0.42–1)

PPV = 0.92

NPV = 0.75

p-Value = 0.000

Neural net (ROC)

AUC = 0.76 (0.54–0.98)

sensitivity = 0.54 (0.25–0.81)

specificity = 1 (0.59–1)

PPV = 1

NPV = 0.54

p-Value = 0.028

Celiac vs. healthy 100 features

Sparse logistic regression (ROC)

AUC = 0.57 (0.34–0.79)

sensitivity = 0.38 (0.14–0.68)

specificity = 0.94 (0.71–1)

PPV = 0.83

NPV = 0.67

p-Value = 0.268

Random forest (ROC)

AUC = 0.51 (0.28 - 0.75)

sensitivity = 0.38 (0.14 - 0.68)

specificity = 0.88 (0.64 - 0.99)

PPV = 0.71

NPV = 0.65

p-Value = 0.442

Gaussian process (ROC)

AUC = 0.5 (0.26–0.74)

sensitivity = 0.23 (0.05–0.54)

specificity = 1 (0.8–1)

PPV = 1

NPV = 0.63

p-Value = 0.5

Support vector machine (ROC)

AUC = 0.71 (0.51–0.91)

sensitivity = 0.92 (0.64–1)

specificity = 0.65 (0.38–0.86)

PPV = 0.67

NPV = 0.92

p-Value = 0.024

Neural net (ROC)

AUC = 0.62 (0.39–0.85)

sensitivity = 0.54 (0.25–0.81)

specificity = 0.88 (0.64–0.99)

PPV = 0.78

NPV = 0.71

p-Value = 0.129

Refractory vs. healthy 100 features

Sparse logistic regression (ROC)

AUC = 0.42 (0.13–0.71)

sensitivity = 0.82 (0.57–0.96)

specificity = 0.29 (0.037–0.71)

PPV = 0.74

NPV = 0.4

p-Value = 0.284

Random forest (ROC)

AUC = 0.42 (0.14–0.7)

sensitivity = 0.59 (0.33–0.82)

specificity = 0.57 (0.18–0.9)

PPV = 0.77

NPV = 0.36

p-Value = 0.2834

Gaussian process (ROC)

AUC = 0.59 (0.29–0.89)

sensitivity = 0.82 (0.57–0.96)

specificity = 0.57 (0.18–0.9)

PPV = 0.82

NPV = 0.57

p-Value = 0.267

Support vector machine (ROC)

AUC = 0.57 (0.29–0.86)

sensitivity = 0.71 (0.44–0.9)

specificity = 0.57 (0.18–0.9)

PPV = 0.8

NPV = 0.44

p-Value = 0.310

Neural net (ROC)

AUC = 0.6 (0.35–0.84)

sensitivity = 0.35 (0.14–0.62)

specificity = 1 (0.59–1)

PPV = 1

NPV = 0.39

p-Value = 0.772