This paper addresses the problem of identifying differentially expressed genes using false discovery rate (FDR) controlling procedures in the context of DNA microarray data. The authors propose three resampling-based FDR controlling procedures that account for the distribution of test statistics and compare their performance to the naive application of the linear step-up procedure (Benjamini and Hochberg, 1995). The procedures are evaluated using simulated microarray data, and their performance is examined relative to their ease of implementation. The study shows that all four FDR controlling procedures control the FDR at the desired level and retain substantially more power than family-wise error rate (FWE) controlling procedures. Resampling of the marginal distribution of each test statistic improves performance over the naive approach. The highest power is achieved by resampling-based procedures that resample the joint distribution of the test statistics and estimate the level of FDR control. The paper discusses the FDR criterion, the linear step-up procedure (BH), and adaptive procedures that estimate the number of true null hypotheses (m₀) to gain more power. It also presents methods for multiplicity adjusted p-values and resampling FDR adjustments. The results show that FDR controlling procedures are more powerful than FWE controlling procedures, and that resampling-based FDR adjustments provide better performance. The study includes a comparative simulation study where the performance of the multiple testing procedures is evaluated under controlled differential expression. The results show that all FDR controlling procedures produce adjusted p-values much closer to the true FDR than the FWE adjusted p-values obtained by the WY algorithm. The resampling point-estimator is the most powerful procedure, followed closely by the other two resampling estimators. The paper concludes that controlling the FDR is more appropriate for gene expression analysis than controlling the FWE, as it allows for more powerful procedures. The FDR criterion is economically interpretable, as it gives the proportion of the investment that is about to be wasted on false leads. The choice among the four procedures is a matter of buying more power and better properties at the expense of more complicated computations. The authors emphasize that resampling-based FDR adjustments provide substantial gains in power and are more effective than the naive approach. The resampling upper limit estimator offers both FDR control and a control on the empirical FDR level.This paper addresses the problem of identifying differentially expressed genes using false discovery rate (FDR) controlling procedures in the context of DNA microarray data. The authors propose three resampling-based FDR controlling procedures that account for the distribution of test statistics and compare their performance to the naive application of the linear step-up procedure (Benjamini and Hochberg, 1995). The procedures are evaluated using simulated microarray data, and their performance is examined relative to their ease of implementation. The study shows that all four FDR controlling procedures control the FDR at the desired level and retain substantially more power than family-wise error rate (FWE) controlling procedures. Resampling of the marginal distribution of each test statistic improves performance over the naive approach. The highest power is achieved by resampling-based procedures that resample the joint distribution of the test statistics and estimate the level of FDR control. The paper discusses the FDR criterion, the linear step-up procedure (BH), and adaptive procedures that estimate the number of true null hypotheses (m₀) to gain more power. It also presents methods for multiplicity adjusted p-values and resampling FDR adjustments. The results show that FDR controlling procedures are more powerful than FWE controlling procedures, and that resampling-based FDR adjustments provide better performance. The study includes a comparative simulation study where the performance of the multiple testing procedures is evaluated under controlled differential expression. The results show that all FDR controlling procedures produce adjusted p-values much closer to the true FDR than the FWE adjusted p-values obtained by the WY algorithm. The resampling point-estimator is the most powerful procedure, followed closely by the other two resampling estimators. The paper concludes that controlling the FDR is more appropriate for gene expression analysis than controlling the FWE, as it allows for more powerful procedures. The FDR criterion is economically interpretable, as it gives the proportion of the investment that is about to be wasted on false leads. The choice among the four procedures is a matter of buying more power and better properties at the expense of more complicated computations. The authors emphasize that resampling-based FDR adjustments provide substantial gains in power and are more effective than the naive approach. The resampling upper limit estimator offers both FDR control and a control on the empirical FDR level.