Table 1 ANN performance in identification driver/non-driver genes of different cancer types.

From: Ontology-based prediction of cancer driver genes

Experiments

F-score

AUC

Number of annotated genes

CMPO

78.95%

72.96%

13,116

GO

84.23%

78.81%

17,591

MP

85.39%

79.51%

7,884

CMPO + GO + MP (Intersection)

91.20%

89.73%

7,884

CMPO + GO + MP (Union)

92.57%

94.28%

20,352

  1. We first determine the performance using each ontology-based feature separately (using each ontology individually as background knowledge), and then determine the performance only on genes for which we have all three ontology-based features (Intersection); in the “Intersection” case, embedding vectors are concatenated. Finally, we determine the performance on genes for which we have at least one feature (Union); in the latter case, ontologies are merged.
Back to article page