Mapping biological pathways across microbial genomes is a highly important technique in functional studies of biological systems. Existing methods mainly rely on sequence-based orthologous gene mapping, which often leads to suboptimal mapping results because sequence-similarity information alone does not contain sufficient information for accurate identification of orthology relationship. Here we present an algorithm for pathway mapping across microbial genomes. The algorithm takes into account both sequence similarity and genomic structure information such as operons and regulons. One basic premise of our approach is that a microbial pathway could generally be decomposed into a few operons or regulons. We formulated the pathway-mapping problem to map genes across genomes to maximize their sequence similarity under the constraint that the mapped genes be grouped into a few operons, preferably coregulated in the target genome. We have developed an integer-programming algorithm for solving this constrained optimization problem and implemented the algorithm as a computer software program, P-MAP. We have tested P-MAP on a number of known homologous pathways. We conclude that using genomic structure information as constraints could greatly improve the pathway-mapping accuracy over methods that use sequence-similarity information alone.genomic structure ͉ operon ͉ regulon ͉ ortholog ͉ pathway mapping C omparative genome analysis represents a powerful technique for functional inference of genes. Its foundation is the ability to identify homologous (or more specifically, orthologous) genes. Here orthologous genes refer to isofunctional and heterospecic genes (1-3) for practical purposes. Several methods have been developed for mapping orthologous genes through sequence comparison. A popular approach is based on reciprocal BLAST searches, the so-called bidirectional best-hit (BDBH) approach (4). Based on this strategy, Wall et al. (5) recently designed a more sophisticated scheme for finding orthologous genes through BLAST searching followed by a more accurate sequence-alignment scheme (e.g., CLUSTALW and PAML). Koonin and co-workers (6) developed a popular method for orthologous gene identification based on the idea of clusters of orthologs groups (COG). Whereas all these approaches have provided useful practical tools for prediction of orthologous genes, their prediction accuracy has been less than optimal, as discussed below. One of the key issues with all these methods is their underlying assumption that sequence similarity alone contains sufficient information for prediction of orthologous gene relationship, which is probably far from being true.One important piece of information for orthologous gene identification across microbial genomes could come from the genomic structures such as operons (7) and regulons (8). Using such information has proven to be very helpful in orthology mapping in our previous studies (9, 10). Based on this observation, we have developed an algorithm, called P-MAP, for mapping orthologous...