A statistical approach in using cDNA array analysis to determine modest changes in gene expression in several brain regions after neurotoxic insult

Modest changes in gene expression of three-fold or less might be expected after mild to moderate neurotoxic exposure to classes of compounds, such as the substituted amphetamines, or at time points that are weeks after more severe neurotoxic exposures. When many genes appear to change expression by less than two-fold, it is crucial to run several pairs of arrays and use statistical analysis to determine which genes are really changing. This limits the number of genes that have to undergo the time consuming task of performing RT-PCR to validate change in expression levels. We describe here methods for statistically determining which genes are being expressed above background levels. These methods are used to compare expression differences among the striatum, parietal cortex, posterior lateral amygdaloid nucleus, and substantia nigra brain regions, all of which differ significantly in their gene expression profiles. In these comparisons, it was possible to distinguish differences among hundreds of genes with manageable estimated false discovery rates. The effect of amphetamine treatment on gene expression in posterior lateral amygdaloid nucleus was also evaluated. The expression data indicate that many genes have changed, but in this case it is more difficult to separate affected genes from false positives. The optimum list has 50 genes, of which 32% are expected to be false positives.