Table 1.

The summary of methods for identifying biclusters with constant row or column on the gene expression datasets.

Algorithm	Methodology	Description
BiMax [11]	Seeks the rectangles of ‘1’′s in a binary matrix	Only suitable for the bicluster with constant up-regulated condition; sensitive to the noise and number of biclusters; affected by the overlap
Plaid [10]	Assume the bicluster is generated as the sum of a background effect, cluster effects, row effects, column effects and random noise	Both suitable for conditions of the bicluster with constant value and constant row/column; sensitive to the noise; affected by the overlap
Spectral [9]	Advantages over SVD spectral analysis of the original or rescaling raw data	Both suitable for conditions of the bicluster with constant up- or down- regulated condition; not sensitive to the noise; not suitable for the discrete datasets; limited in running speed on large datasets; affected by the overlap
Xmotifs [12]	A nondeterministic greedy algorithm that seeks biclusters with conserved rows/columns	Only suitable for conditions of the bicluster with constant row/column; required for dataset discretized and more sensitive to the noise; affected by the overlap limited in running speed on large datasets; affected by the overlap
FABIA [16]	Analysis for bicluster acquisition models the data matrix as the sum of biclusters plus additive noise, bicluster is the outer product of two sparse vectors	Both suitable for conditions of the bicluster with constant value and constant row/column; not sensitive to the noise and the number of biclusters; affected by the overlap
ISA [14]	A nondeterministic greedy algorithm that seeks biclusters from starting with a seed bicluster and re-running the iteration steps	Both suitable for conditions of the bicluster with constant value and constant row/column; not sensitive to the noise the number of biclusters, and the overlaps; limited in running speed on large datasets