Table 1

Accuracy and stability results using different clustering programs, initialization schemes and number of genes/clusters.

Clusters

5000 genes

10000 genes

Combined

Accuracy

Cluster*

4

0.405 (0.404–0.405)

0.594 (0.569–0.604)

0.487 (0.404–0.604)

PKM-RIA*

4

0.519 (0.453–0.597)

0.896 (0.896–0.896)

0.747 (0.453–0.896)

PKM-RFD*

4

0.519 (0.322–0.519)

0.770 (0.586–0.896)

0.553 (0.322–0.896)

PKM-BKM*

4

0.489 (0.489–0.489)

0.770 (0.770–0.770)

0.629 (0.629–0.629)

Cluster*

20

0.163 (0.124–0.183)

0.256 (0.196–0.297)

0.190 (0.124–0.297)

PKM-RIA*

20

0.231 (0.211–0.461)

0.216 (0.208–0.233)

0.227 (0.208–0.461)

PKM-RFD*

20

0.189 (0.178–0.226)

0.210 (0.202–0.252)

0.202 (0.178–0.252)

PKM-BKM*

20

0.400 (0.400–0.400

0.691 (0.691–0.691)

0.545 (0.400–0.691)

Stability

Cluster*

4

0.439 (0.436–0.442)

0.797 (0.783–0.812)

0.618 (0.436–0.812)

PKM-RIA*

4

0.514 (0.435–1.00)

1.00 (1.00–1.00)

1.00 (0.435–1.00)

PKM-RFD*

4

0.569 (0.483–1.00)

0.769 (0.586–0.770)

0.711 (0.483–1.00)

PKM-BKM*

4

1.00 (1.00–1.00)

1.00 (1.00–1.00)

1.00 (1.00–1.00)

Cluster*

20

0.347 (0.321–0.393)

0.492 (0.418–0.907)

0.405 (0.321–0.907)

PKM-RIA*

20

0.738 (0.444–0.888)

0.904 (0.682–0.994)

0.788 (0.444–0.994)

PKM-RFD*

20

0.594 (0.548–0.643)

0.652 (0.573–0.668)

0.634 (0.548–0.668)

PKM-BKM*

20

1.00 (1.00–1.00)

1.00 (1.00–1.00)

1.00 (1.00–1.00)

* – Cluster k-means: Single Node (N = 12), All PKM: 1 and 7 node data (N = 12)

All values are the median adjusted Rand Index with the range of values in parentheses.

Accuracy = cluster results vs. known assignments, Stability = over all agreement between cluster results

PKM = ParaKMeans, RFD = Random From Data, RIA = Random Initial Assignment, BKM = Bissecting K Means, Cluster = Eisen Cluster program.

Kraj et al. BMC Bioinformatics 2008 9:200   doi:10.1186/1471-2105-9-200