Comparison of consistency across sub data sets and consistency across redundant probe sets based on re-sampling.

The subset of data corresponding to low reaction level were divided into two data sets by randomly sampling 3 out of the 6 biological replicates from each brain region. Each probe set was individually and with its redundant probe set partner for differential expression, fold-change correlations across and within sub-data sets, as described in the Methods. The mean and standard deviation were calculated for each summary statistics across twenty re-samplings of the data.