Table 7.

Statistics for comparing the performance of the tested biclustering algorithms with different preprocessing techniques. Note: $|\mathcal{B}|$ corresponds to the number of biclusters; $\overline{|I|}$ is the average number of genes per bicluster; σ_|I|the standard deviation of genes per bicluster; $\overline{|J|}$ the average number of conditions per bicluster; σ_|J| the standard deviation of the number of conditions per bicluster; and finally $\overline{\text{Terms}}$ the average number of enriched terms per bicluster.

Algorithm	Preprocessing	$|\mathcal{B}|$	$\overline{|I|}$	σ|I|	$\overline{|J|}$	σ|J|	$\overline{\text{Terms}}$
BicPAMS	p < 0.01	80	208.03	18.54	3.16	0.53	28.91
	p < 0.05	79	3526.66	301.50	3.24	0.64	341.70
	ANOVA (top 200)	7	188.29	5.95	10.00	9.70	10.57
	ANOVA (top 1000)	20	676.05	29.13	5.00	4.22	55.75
	ANOVA (top 5000)	57	2106.18	128.36	3.61	1.25	131.32
Cheng and Church	p < 0.01	50	15.60	12.59	12.92	5.90	3.46
	p < 0.05	100	55.90	16.23	34.79	9.96	1.68
	ANOVA (top 200)	8	25.00	23.49	21.38	12.56	6.50
	ANOVA (top 1000)	56	17.86	15.27	17.89	10.67	4.41
	ANOVA (top 5000)	100	34.54	24.10	22.76	11.25	2.47
Plaid	p < 0.01	10	64.70	55.53	14.20	5.60	29.90
	p < 0.05	10	776.40	922.43	11.60	6.89	24.10
	ANOVA (top 200)	8	44.00	30.76	12.88	8.43	9.88
	ANOVA (top 1000)	10	159.50	100.72	12.20	7.29	18.70
	ANOVA (top 5000)	10	739.20	530.04	13.10	7.48	43.40
xMotifs	p < 0.01	10	31.90	17.17	8.20	2.86	1.70
	p < 0.05	10	654.50	365.82	6.00	0.00	6.30
	ANOVA (top 200)	6	30.33	34.30	24.50	9.73	10.67
	ANOVA (top 1000)	10	71.90	103.54	11.10	4.28	7.60
	ANOVA (top 5000)	10	326.00	538.95	6.20	0.60	5.30