Supplementary Information for:

"Feature  Construction from synergic pairs to improve microarray-based  classification"

Blaise Hanczar, Jean-Daniel Zucker, Corneliu Henegar, Lorenza Saitta

Table1:
Comparison of the impact of features selection methods on classification accuracy. The classification error rate is estimated by bootstrap632. Table1

Table2:
Comparison of the impact of features selection methods on classification accuracy. The classification error rate is estimated by cross-validation. Table2

Figure1:
Representation of the experimental design to compute the classification error rate. Figure1

Figure2:
Comparison of the information of the genes with the information of the pairs of genes. Figure2

Figure3:
Study of the ranks of genes forming the best pairs of genes. Figure3

Figure4:
Comparison of the information of the constructed features and the information of the genes. Figure4

Accuracy of the mutual information measures:
Study of the accuracy of the mutual information measures. Results