Because a very large number of gene expression data sets are currently publicly available, comparisons across experiments between different laboratories have become a common task. However, most existing methods of comparing gene expression data sets require setting arbitrary cutoffs (e.g., for statistical significance or fold change), which could select genes according to different criteria because of differences in experimental protocols and statistical analysis in different data sets. A new method is proposed for comparing expression profiles across experiments by using the rank of genes in the different datasets. We introduce a maximization statistic, which can be calculated recursively and allows for efficient searches on a large space (paths on a grid). We apply our method to both simulated and real datasets and show that it outperforms other existing rank-based algorithms. CORaL is a novel method for comparison of gene expression data that performs well on simulated and real data. It has the potential for wide and effective use in computational biology.