There are several different algorithms published for the identification of differentially expressed genes in DNA microarray experiments. Such algorithms produce ordered lists of genes. To compare the performance of these algorithms established measurements from Information Retrieval are proposed. A benchmark data set with known properties is generated and published. This benchmark data is used to compare the performance of different algorithms with a new algorithm, called PUL. Surprisingly a clear ordering in performance of the algorithms was observed. PUL outperformed other algorithms by a factor of two. PUL was applied successfully in different practical applications. For these experiments the importance of the genes identified by PUL were independently verified.