ABSTRACT
Each gene’s differential expression pattern in a microarray experiment is usually assessed by (typically pairwise) contrasts of mean expression values among experimental conditions. Such comparisons can be made as fold changes whereby genes with greater than two-or three-fold changes are selected for further investigation. However, it has been frequently found that a gene showing a high fold change between experimental conditions might also exhibit high variability and hence its differential expression may not be significant. Similarly, a modest change in gene expression may be significant if its differential expression pattern is highly reproducible. A number of authors have pointed out this fundamental flaw in the fold change-based approach (e.g., Jin et al., 2001). And, in order to assess differential expression in a way that controls both false positives and false negatives, a standard approach is emerging as one based on statistical significance and hypothesis testing, with careful attention paid to reliability of variance estimates and multiple comparison issues.
