Composition of the classifier derived from class prediction (sorted by t -value): For feature selection the “univariate p-value <0.05 and 2 fold -change between classes” was applied. The column

The column “%CV-support” in the table indicates the percentage of the cross-validation training sets in which each gene was selected during “leave-one out cross validation”. 100% means that the gene is so strong that it was selected in all of the cross-validated training sets. “Geom mean of intensities” is derived from transformed qPCR-Ct values.