Skip to main content

Table 4 Comparison of cancer driver variant predictors on the DRGN benchmark, with weighting of the predictions of the variants on each gene in function of the likelihood of Gi to host predominantly positive or negative samples. The last column shows the AUC drop with respect the EXPDRV+ALL data set in Table 2

From: Current cancer driver variant predictors learn to recognize driver genes instead of functional variants

Method

AUC

AUPRC

AUC drop

ParsSNP

52.9

52.3

34.0

CanDrA v+

51.7

52.1

35.4

Chasm 3.1

64.5

61.4

26.6

CHASMplus

59.2

57.9

35.1

TransFIC

57.5

54.8

10.5

FATHMM Cancer

57.6

54.7

33.4

DummyLogOdd

50.0

50.0

48.1