We computed Partial Least Squares Discriminant Analysis (PLS-DA) models for each isolation method to examine sets of genes whose transcripts are responsible for separating the methods. For each model, we extracted the variable weights of the expression probes, ranked these variable weights and selected the 5% highest and 5% lowest ranked expression probes. Table 2 shows genes that were strongly up- or down-regulated in the PLS-DA models.
