The Optimal Discovery Procedure (ODP) is a new method that finds the optimal rule for calling genes differentially expressed. This is a theoretically sound method proven in general that it cannot be outperformed...
Using new statistical theory (Storey 2005, Storey, Dai and Leek 2005), EDGE allows one to identify genes that are differentially expressed between two or more different biological conditions (e.g., healthy versus diseased tissue). There are already a number of software packages to perform this type of analysis. However, EDGE is based on the Optimal Discovery Procedure (ODP) that we recently developed, which is significantly different from existing approaches. Whereas previously existing methods employ statistics that are essentially designed for testing one gene at a time (e.g., t-statistics and F-statistics), the ODP uses all relevant information from all genes in order to test each one for differential expression. The improvements in power are substantial; the figure shows a comparison between EDGE and the five most cited methods previously available based on a well-known breast cancer expression study (Hedenfalk et al. 2001). It can be seen that our ODP approach increases the number of genes called significant by 82% on average in comparison to these methods. For example, at a false discovery rate cut-off of 3%, EDGE finds 136 significant genes while the highly popular SAM software only finds 30 genes. More details on this method will be posted here when the papers are published.
A comparison between EDGE and the top five most cited procedures for identifying differentially expression genes applied to the BRCA data set of Hedenfalk et al. 2001. For each q-value (false discovery rate) cut-off, the number of genes found to be significantly differentially expressed between BRCA1 and BRCA2 mutation-positive tumors is plotted for each procedure.