Table 8.

Comparison of the clustering performance among GO Fuzzy c-means, FuzzyK, SOM and Gaussian mixture model using datasets A and B. GOFuzzy_x% represents x percentage of GO annotation was used in GO Fuzzy c-means. ClusterJudge [33] was used to compute the z-scores with 10 runs for each of the clustering results. A clustering result with higher z-score indicates that the clusters are more likely to be biologically relevant.

Method	z-scores and standard error for Dataset A	z-scores and standard error for Dataset B
GOFuzzy25%	91.18 ± 3.22	119.10 ± 3.47
GOFuzzy50%	175.90 ± 4.68	181.20 ± 3.36
GOFuzzy75%	248.40 ± 3.81	255.10 ± 4.65
GOFuzzy100%	323.10 ± 7.59	316.60 ± 6.04
FuzzyK	102.33 ± 1.85	108.10 ± 2.32
Fuzzy c-means	68.08 ± 5.52	83.12 ± 4.57
FuzzySOM	68.56 ± 2.66	81.48 ± 5.43
FLAME	66.18 ± 4.83	85.55 ± 5.93
SOM	44.13 ± 0.61	52.62 ± 0.30
Gaussian	0.72 ± 0.030	73.55 ± 0.77