Supplement material for Figure 1 and Figure 2:

• Tables with (1) the best cR for the Affymetrix data sets for each clustering algorithm and (2) difference between the actual number of classes and the number of clusters in the partition solutions with best cR.

• Pearson Correlation
• Cosine
• Spearman Coefficient
• Euclidean Distance with Z_0
• Euclidean Distance with Z_1
• Euclidean Distance with Z_2
• Euclidean Distance with Z_3

Supplement material for Figure 3 and Figure 4:

• Tables with (1) the best cR for the cDNA data sets for each clustering algorithm and (2) difference between the actual number of classes and the number of clusters in the partition solutions with best cR.

• Pearson Correlation
• Cosine
• Spearman Coefficient
• Euclidean Distance with Z_0
• Euclidean Distance with Z_1
• Euclidean Distance with Z_2
• Euclidean Distance with Z_3

Supplement material for the paired t-test:

We calculate the mean of the cR over all data sets in two different contexts: (A) taking into account, for each data set, only the

partition with the number of clusters equals to the number of actual classes in the respective data set, and (B) considering, for each data set, the partition that presents the best cR, independently of its number of cluster. In the case of the data for the paired t-test to check if differences in the cR between distinct method are statistical significant, given that we are mainly interested in comparing the clustering methods, we selected for each context, only the results with the proximity measure, which displayed the largest mean of cR.

• Affymetrix
• Context A
• Context B
• cDNA
• Context A
• Context B