Material to A note on oligonucleotide expression values not being normally distributed

Motivation: Novel techniques for analyzing microarray data are constantly being developed. Though many of the methods contribute to biological discoveries, inability to properly evaluate the novel techniques limits their ability to advance science. Because the underlying structure, or distribution, of microarray data is unknown, novel methods are typically tested against the assumed structure of normally distributed data. However, microarray data are not, in fact, normally distributed, and testing against such data can have misleading consequences. Results: Using an Affymetrix technical replicate Spike-In data set, we showed that oligonucleotide expression values are not universally normally distributed under any of the standard methods for extracting expression values. The resulting data tend to have a large proportion of skew and heavy tailed values. Using data simulated under three models (normal, heavy tailed, and skewed), additionally, we showed that standard methodologies (for differential expression and gene similarity) can give unexpected and misleading results when the data are not normally distributed. Robust methods should be used when analyzing microarray data. Additionally, when evaluating new techniques, skewed and/or heavy tailed data distributions should be considered in simulations.