Table 4

Combinations of GO attributes: Shown is a schematic example of 8 genes, separated into two distinct clusters (Table 4a). As can be observed neither of the two given attributes is significantly enriched within any of the cluster, resulting in a vanishing mutual information between the clustering and the annotation (see the respective contingency tables in Table 4b and 4c). However, clearly the combination of both attributes does uniquely determine both cluster. In particular, genes with the combination A = (A1, A2) = (0, 0) or (1, 1) are grouped together in the first cluster 0. while genes sharing the annotation A = (0, 1) or (1, 0) are grouped together in the second cluster.

4a

gene

1

2

3

4

5

6

7

8


cluster

0

0

1

0

1

1

1

0


A1

1

0

0

0

0

1

1

1

A2

1

0

1

0

1

0

0

1


4b

Cluster:

0

1


A1

0

2

2

1

2

2


4c

Cluster:

0

1


A2

0

2

2

1

2

2


Steuer et al. BMC Bioinformatics 2006 7:380   doi:10.1186/1471-2105-7-380

Open Data