Skip to main content

Table 5 Results for composite predictive models. The method for combination is displayed in the first column, the estimate of AUC obtained on the training set and in cross-validation in the second and third column. Columns four and five present results of comparison between the predictions of composite models and the best individual model i.e. MCF-7+molecular descriptors, by paired t-test

From: Integration of human cell lines gene expression and chemical properties of drugs for Drug Induced Liver Injury prediction

 

AUC

Method

  

Comparison with MCF-7+chem.

 

Internal CV

Nested CV

t statistic

p-value

Best single result

0.69

0.65

-6.8

1

Mean of all results

0.66

0.67

-7.0

1

Mean of 5 best results

0.72

0.73

6.8

9·10−7

Linear combination

0.73

0.70

-0.46

0.67

Random Forest

0.84

0.71

1.8

0.05