Skip to main content

Table 3 Comparing biclustering algorithms on the basis of their predictive capacities for recovering selected patterns.

From: Construction of gene regulatory networks using biclustering and bayesian networks

GO Term/number of annotated genes

K-means

CC

ISA

Bivisu

OPSM

GO:0042493

Response to drug (118)

4

5

7

6

0

GO:0006970

Response to osmotic stress (83)

3

5

6

3

0

GO:0006979

Response to oxidative stress (79)

2

7

11

0

0

GO:0046686

Response to cadmium ion (102)

2

3

2

2

0

GO:0043330

Response to exogenous dsRNA (7)

2

3

2

2

0

GO:0046685

Response to arsenic (77)

2

0

2

2

0

GO:0006950

Response to stress (532)

9

11

16

2

0

GO:0009408

Response to heat (24)

3

0

2

2

0

GO:0009409

Response to cold (7)

0

0

2

0

0

GO:0009267

Cellular response to starvation (44)

0

2

0

0

0

GO:0006995

Cellular response to nitrogen starvation (5)

4

4

4

0

0

GO:0042149

Cellular response to glucose starvation (5)

0

2

0

0

0

GO:0009651

Response to salt stress (15)

2

7

0

0

0

GO:0042542

Response to hydrogen peroxide (5)

0

0

0

2

0

GO:0006974

Response to DNA damage stimulus (240)

0

22

0

3

0

GO:0000304

Response to singlet oxygen (4)

2

0

0

0

0

  1. We tested the predictive capacities of different biclustering algorithms for recovering the gene ontology category within response to stress (GO:0006950). Rows represent the known gene ontology function categories under the response to stress category and the columns are tge different biclustering methods, with the highest relevant performance cluster result as the entry for a given functional category and clustering method. Several interesting observations can be made. First, although OPSM showed a high percentage of enriched biclusters, it had no biclusters with genes matching any of the known GO categories for the Gasch data set. Second, despite the low number of biclusters (9) and low percentage of gene coverage (25\%), ISA showed better performance, with one of its biclusters having 11 genes matching response to oxidative stress (GO:0006979). Finally, we observe that several methods appear to be unique in detecting biclusters related to certain function categories. For example, ISA and CC detected two genes belonging to the response to cold and cellular response to starvation functions, respectively.