Reproducibility of microarray results (updated 12 Jan 01)

We and others have found that the amount of random error associated with microarrays appears to be higher than was originally expected. However, this error can be overcome by increasing the number of replicate spots printed on each slide, and by increasing the number of replicate hybridizations. We have obtained our most reproducible results, in terms of coefficient of variation of signal intensity, by printing spots in replicates of 5, and discarding the highest and lowest spots from each set prioir to analysis (see figure below).