PMC

The first two principal components of the 157-gene model in the Swedish and Physicians' Health Study data sets, comparing men with Gleason ≤ 6 (Both ≤ 3) and men with Gleason ≥ 8 (Both ≥ 4).

Misclassification error rates from prediction analysis of microarrays estimated by cross validation in the Swedish cohort for models with differing numbers of genes. These models try to distinguish between men with Gleason ≤ 6 and those with Gleason ≥ 8. The red line is the proportion of samples misclassified, with error bars given in black.

Receiver operating characteristic (ROC) curves for the 157-gene signature developed with prediction analysis of microarrays. (A) ROC curve for the model built in the Swedish cohort for Gleason ≤ 6 and Gleason ≥ 8, applied to those same men. (B) ROC curve for the model built in the Swedish cohort applied to men in the Physicians' Health Study with Gleason ≤ 6 and Gleason ≥ 8. In each plot, the gold dot is where misclassification rates are approximately equal in the two groups.