Table 2.

Protein interaction modules predicted to be the most discriminating markers of cancer phenotype.

Gene module					N0	N1	H	Prediction accuracy
Id	Gene1	Gene2	Gene1 state	Gene2 state
1	CANX	FAM107A	0	1	27	1	0.222285	0.94
2	ABCB1	CAV1	1	1	27	1	0.222285	1.00
3	COL10A1	P4HB	0	0	27	1	0.222285	0.98
4	PAICS	CHD3	0	0	26	1	0.228538	0.98
5	CAV1	SRC	1	1	26	1	0.228538	0.89
6	TNFRSF1B	SSR4	1	0	26	1	0.228538	0.91
7	LMO2	MAPRE3	1	0	26	1	0.228538	0.96
8	SMAD3	EPAS1	0	1	26	1	0.228538	0.94
9	NOS3	CAV1	0	1	26	1	0.228538	0.94
10	COL10A1	P4HB	0	0	26	1	0.228538	0.96
11	PDK1	EPAS1	0	1	26	1	0.228538	0.96
12	LMO4	TCF21	0	1	26	1	0.228538	0.96
13	SKIL	SASH1	1	1	26	1	0.228538	0.96
14	PAFAH1B1	NDEL1	0	0	1	26	0.228538	0.89
15	CAV1	SRC	0	1	1	26	0.228538	0.89
16	NOS3	CAV1	0	0	1	26	0.228538	0.94

N₀, number of times that the state S appeared in normal samples. N₁, the number of times it appeared in cancerous samples. H, Calculated Entropy, Prediction Accuracy is calculated applying Leave-one-out cross-validation on SVM classifier (see Materials and Methods for more details).