1: The number of misclassified samples. 2: Sets of genes (probe sets) produced by iteratively removing 10% of the least significant genes at a time. 3: D1: Data set 1; D2: Data set 2; C: Combined data set; D1→D2: use D1 to predict D2; D2→D1: use D2 to predict D1. +: Additional probe set ID plus those identified from lower number probe-set list(s). For example, the 3 probe sets contains 1596_g_at in addition to those (35868_at and 268_at) identified from the 2 probe sets prediction.