Fig. 4

Performances of methods for protein level prediction. Boxplots show the distribution of the normalized root mean square error (NRMSE) in the prediction of protein levels, using six described methods: Averaged Protein (AP), Weighted Average Protein (WAP), Average PTR (APTR), Weighted Average PTR (WAPTR), FC Based (FCB), and Relaxed FCB (RFCB). In each tissue, RMSE values are divided by the standard deviation of the protein levels in that tissue. The error sizes are averages over tissues of 10-fold cross validation. In the EAR dataset there are only two groups, so the weighted/relaxed versions are irrelevant. Boxplots show median, a box for the middle 50% and whiskers to the largest and smallest values that are not classified as outliers. If the distance of an observation from the box is higher than 1.5 times the box size, it is classified as an outlier. Outliers are labeled