Representational difference analysis (RDA) is a powerful and sensitive tool for identification of differentially expressed genes (M. Hubank and D. G. Schatz, 1999, Methods Enzymol. 303, 325-349; 1994, Nucleic Acids Res. 22, 5640-5648) that will identify both up- and downregulated genes differentially expressed between two cDNA populations. This manuscript provides a thorough description of an optimized RDA method. This procedure while still based on the traditional RDA originally developed by Lisitsyn and co-workers(N. A. Lisitsyn, 1995, Trends Genet. 11, 303-307; N. A. Lisitsyn, F. S. Leach, B. Vogelstein, and M. H. Wigler, 1994, Cold Spring Harbor Symp. Quant. BioL 59, 585-587; N. Lisitsyn, N. Lisitsyn, and M. Wigler, 1993, 259, 946-951) and modified by Hubank and Schatz for RNA (1994, Nucleic Acids Res. 22, 5640-5648) is improved and requires less starting material than many existing methods. Several key modifications are included (1). Size-exclusion gel-filtration microspin columns are used throughout the procedure to remove the primers and low molecular weight cDNAs. This results in reducing the number of ethanol precipitations required and in improving the yield of desirable amplification products (2). Elimination of the mung bean nuclease treatment in favor of a simple dilution of PCR serves as a means of markedly reducing the single-stranded cDNAs that can interfere with the amplification of differentially expressed products (3). The use of up to six unique noninteracting primers ensures that no anomalous amplification occurs due to carryover of primers or incomplete digestion from the ends of the cDNAs (4). A set of cDNA standards was developed and various concentrations were used to better characterize the ability of representational difference analysis to identify rare messages in a complex cDNA population (5). Integral to this manuscript, a detailed laboratory protocol is available from the authors (craig.byus@ucr.edu) and provides a step-by-step description of the modified procedure.