A probe designed against a landmark gene was progressed if its z-score when targeted by an shRNA was -2.0 or lower. When the initial probe design showed non-specific reactivity, failed to correlate with reference mRNA standards or failed to register adequate knockdown, we redesigned the probe sequence and retested. After a few cycles of iteration between design and empirical testing, we were able to show that 846 of the 955 targeted landmark genes (89%) were down-regulated by at least one targeting shRNA (z-scores less than -2). However, a low z-score doesn't in itself imply specificity—for example, a sample corrupted by dead cells might have yielded low mRNA across the board, leading to many genes with low z-scores. To guard against nonspecific reduction of z-scores, we compared the distribution of targeted gene z-scores to non-targeted gene z-scores and observed that the former was significantly left-shifted, indicating that the observed down-regulation is largely specific to the targeted genes (Figure 1C, middle panel). For each targeted gene, we computed the rank of its z-score in the experiment in which it was targeted relative