UNCLES: method for the identification of genes differentially consistently co-expressed in a specific subset of datasets

Table 3 Comparison between UNCLES and eight biclustering methods

Cluster and s	UNCLES^ab	Plaid^a	Bimax^a	FLOC^a	LAS^a	BGSP^a
C1 1200	0.00	0.10	1.00	1.35	0.13	0.61
C1 1200	82/82	20/82	4/82	6/82	21/82	1/82
C1 2000	0.00	0.64	1.06	1.38	0.16	0.75
C1 2000	82/82	22/82	4/82	6/82	21/82	2/82
C1 3000	0.00	0.95	1.12	1.39	0.29	0.90
C1 3000	82/82	37/82	4/82	6/82	18/82	0/82
C1 5000	0.04	1.28	1.21	1.40	0.45	0.06
C1 5000	82/82	5/82	3/82	6/82	18/82	0/82
C1 7000	0.02	0.97	0.95	1.40	0.59	0.09
C1 7000	82/82	30/82	4/82	6/82	19/82	0/82
C2 1200	0.00	0.76	1.21	1.36	0.31	0.96
C2 1200	42/42	5/42	3/42	2/42	15/42	0/42
C2 2000	0.00	0.92	1.26	1.37	0.28	0.91
C2 2000	42/42	16/42	3/42	3/42	15/42	0/42
C2 3000	0.33	0.99	1.29	1.38	0.32	1.00
C2 3000	42/42	5/42	3/42	5/42	15/42	0/42
C2 5000	0.40	1.07	1.32	1.40	0.71	1.14
C2 5000	42/42	5/42	3/42	2/42	13/42	0/42
C2 7000	0.43	1.18	1.30	1.40	0.70	1.17
C2 7000	42/42	5/42	3/42	4/42	13/42	0/42

^aEach cell in those columns includes two values – the first is the distance from the top-left corner of the ground-truth-based F-P plots for the best cluster found by each method; the ideal is zero and the maximum is \( \sqrt{2}\cong 1.41 \); the second value is the number of data samples (data matrix columns) which the algorithms correctly found for the corresponding clusters out of the total number of correct samples (82 for type A and 42 for type B)
^bThe number of data matrix columns (samples) are prefixed for UNCLES while being variable for biclustering methods

ISSN: 1471-2105