Reproducibility Assay Experimental Design

In order to measure the reproducibility of our microarrays in a real experimental setup, we have designed a yeast 30K microarray containing 6540 probe sets that survey 6331 unique transcripts. Each probe set contains 4 identical probe replicates that are randomly distributed across the array. We have synthesized nine yeast microarrays (3 batches of 3 slides). Yeast was grown on two different sources of carbon and total RNA were extracted, amplified and labeled with two different fluorophores. A single master hybridization mix was prepared and hybridized to the 9 slides.

For the coefficient of variation (CV) analysis, the ratio (F635/F532) of scaled and background subtracted median signal was used. Spots with signal that did not exceed twice the background in both channels were considered “not present” and eliminated from further analysis. Probe sets qualified for CV analysis if all 4 identical probe replicates within an array had signal, in at least one channel, which exceeded twice the background level in the appropriate channel.

Intra-array Coefficient of Variation

For the coefficient of variation (CV) analysis, the ratio (F635/F532) of scaled and background subtracted median signal was used. Spots with signal that did not exceed twice the background in both channels were considered “not present” and eliminated from further analysis. Probe sets qualified for CV analysis if all 4 identical probe replicates within an array had signal, in at least one channel, which exceeded twice the background level in the appropriate channel.

Green: Batch 1, slides 1 to 3; Yellow: Batch 2, slides 1 to 3; Red: Batch 3, Slides 1 to 3.

Inter-array Coefficient of Variation

Inter-array intra-batch CVs were calculated for each probe if all three probes qualified (signal in at least one channel exceeded twice the background level in the appropriate channel). As seen in the graph below, our inter-array mean CVs are less than 12% (µ = 9.3 ± 2.3). Importantly, no inter-array normalization was performed prior to calculating inter-array CV. These data demonstrate that high signal uniformity is maintained across arrays synthesized contemporaneously.

Inter-batch Coefficient of Variation

Inter-array CV was calculated for each probeset across 9 arrays from 3 independent batches of arrays. Inter-batch CVs were calculated using a variable number (≥4, ≥7, or all 9) of qualifying probesets (see above). As seen in the figure below, inter-batch CVs were less than 10% (µ = 8.8 ± 0.8) and independent of the number of qualifying probe sets. Importantly, no inter-array or inter-batch normalizations were performed. These data show that high signal constancy is preserved across arrays that were manufactured in different batches.