Identification and characterization of functional elements in the noncoding regions of genomes is an elusive and timeconsuming activity whose output does not keep up with the pace of genome sequencing. Hundreds of bacterial genomes lay unexploited in terms of noncoding sequence analysis, although they may conceal a wide diversity of novel RNA genes, riboswitches, or other regulatory elements. We describe a strategy that exploits the entirety of available bacterial genomes to classify all noncoding elements of a selected reference species in a single pass. This method clusters noncoding elements based on their profile of presence among species. Most noncoding RNAs (ncRNAs) display specific signatures that enable their grouping in distinct clusters, away from sequence conservation noise and other elements such as promoters. We submitted 24 ncRNA candidates from Staphylococcus aureus to experimental validation and confirmed the presence of seven novel small RNAs or riboswitches. Besides offering a powerful method for de novo ncRNA identification, the analysis of phylogenetic profiles opens a new path toward the identification of functional relationships between co-evolving coding and noncoding elements.[Supplemental material is available online at www.genome.org.]In all living organisms, the genome regions located between protein-coding sequences are home to a wide diversity of functional elements that include noncoding RNA (ncRNA) genes, DNA regulatory elements, untranslated regions (UTRs) of genes, transposable and self-replicating elements, and a variety of other transcribed or nontranscribed functional sequences. As these elements are often key players in gene regulation and thus in the global cell interaction network, their systematic identification and characterization has become a major challenge in biology.Computational protocols developed to collect and characterize noncoding elements in genomic sequences rely, to a large extent, on comparative genomics. The most common strategies involve, first, collecting sequences under selective pressure and, second, analyzing the aligned sequences using various classifiers that exploit criteria such as nucleotide composition, folding potential, fold conservation, or covariation between distant positions (Rivas and Eddy 2001;Washietl et al. 2005;Pedersen et al. 2006;Torarinsson et al. 2006). In general, such classifiers are designed to detect structured RNAs among noncoding elements with no further distinction between regulatory elements, repeats, or artifacts produced by sequence comparison algorithms.Comparative genomics entails a significant amount of expert intervention, especially in obtaining the right genome set to optimize the specificity and sensitivity of RNA detection. Although a number of studies have successfully identified ncRNAs in several animal (Missal et al. 2005;Washietl et al. 2007) and microbial genomes (Altuvia 2007), the pace at which such studies are performed and published lags far behind the rate of genome sequence output. Most of the complete...