Table 1.

Average sizes of classes compared with random model R

	IN-LSCC	LSCC	OUT-LSCC	SIMPLE	SSCC	INT	EXCP
	Luscombe
actual	9	25	68	38	2	0	2
average	17.1	42.3	43.2	33.8	1.00	2.8	2.8
p-value	0.02	0.025	0.001	0.062	0.6	0.097	0.58
	Yu
actual	20	63	114	77	5	6	5
average	32.5	69.5	102.8	69.6	0.44	6.3	4.2
p-value	0.001	0.002	0.020	0.22	0.01	0.32	0.34
	Balaji
actual	21	60	58	14	0	3	1
average	20.9	74.4	45.6	14.3	0.2	1.2	0.5
p-value	0.53	0.002	0.002	0.57	0.92	0.14	0.35

We use Sscc to denote small cyclic scc's. The TFs that are neither in LSCC nor in its in- or out-components are classified according to MPL, the maximal path length for a path that includes a given TF; when MPL is 1 or 2, TF is in SIMPLE, if MPL is more than 3, TF is in EXCP, and if MPL is 3, we could make either decision, so here we inluded intermediate class INT.

Random graphs were produced to get a uniform distribution among graphs in which TFs have the same in-and out- degrees as in the original network, without changing the TF-TT connections.