Table 3. Classification performance for each data set.

data set	AUC (feat. selection)	AUC (unselected feat.)	MCC (feat. selection)	MCC (unselected feat.)
genes	0.86	0.83	0.55	0.48
domains	0.76	0.82	0.55	0.54
pathways	0.77	0.76	0.37	0.43
PPIs	0.81	0.79	0.44	0.38
genes + domains	0.89	0.87	0.64	0.57
genes + pathways	0.84	0.83	0.5	0.52
genes + PPIs	0.84	0.79	0.49	0.42
domains + pathways	0.88	0.86	0.54	0.58
domains + PPIs	0.87	0.81	0.59	0.45
pathways + PPIs	0.82	0.8	0.47	0.45
genes + domains + pathways	0.89	0.86	0.6	0.56
genes + pathways + PPIs	0.85	0.8	0.53	0.43
genes + domains + PPIs	0.87	0.83	0.56	0.52
domains + pathways + PPIs	0.88	0.82	0.6	0.48
genes + domains + pathways + PPIs	0.86	0.83	0.53	0.52

Performance of prediction models on the test sets averaged on ten repetitions. The highest value of each column is bold. Feature selection increases AUC in 14 out of 15 data sets up to +6%, and MCC in 12 out of 15, up to +12%.