Table 1.

Blind accuracies for the SVM models using different filtering methods

Filtering method	Number of genes	Accuracy (%) SVM model
PART	1000	88.9
NSC	1000	66.7
S2N	1000	77.8
SAM	1000	77.8
Student's t-test	1000	66.7
U-test	1000	66.7
Welch's t-test	1000	66.7
Random selection1	1000	55.6
No filtering	12241	55.6

¹ The SVM model was constructed by using 1000 probes selected randomly. This process was repeated 1000 times. Average accuracies of 1000 SVM models were calculated.