RhoTermPredict: An algorithm for predicting Rho-dependent transcription terminators based on Escherichia coli, Bacillus subtilis and Salmonella enterica databases

Background: In bacterial genomes, there are two mechanisms to terminate the DNA transcription: the "intrinsic" or Rho-independent termination and the Rho-dependent termination. Intrinsic terminators are characterized by a RNA hairpin followed by a run of 6-8 U residues relatively easy to identify using one of the numerous available prediction programs. In contrast, Rho-dependent termination is mediated by the Rho protein factor that, firstly, binds to ribosome-free mRNA in a site characterized by a C > G content and then reaches the RNA polymerase to induce its release. Conversely on intrinsic terminators, the computational prediction of Rho-dependent terminators in prokaryotes is a very difficult problem because the sequence features required for the function of Rho are complex and poorly defined. This is the reason why it still does not exist an exhaustive Rho-dependent terminators prediction program. Results: In this study we introduce RhoTermPredict, the first published algorithm for an exhaustive Rho-dependent terminators prediction in bacterial genomes. RhoTermPredict identifies these elements based on a previously proposed consensus motif common to all Rho-dependent transcription terminators. It essentially searches for a 78 nt long RUT site characterized by a C > G content and with regularly spaced C residues, followed by a putative pause site for the RNA polymerase. We tested RhoTermPredict performances by using available genomic and transcriptomic data of the microorganism Escherichia coli K-12, both in limited-length sequences and in the whole-genome, and available genomic sequences from Bacillus subtilis 168 and Salmonella enterica LT2 genomes. We also estimated the overlap between the predictions of RhoTermPredict and those obtained by the predictor of intrinsic terminators ARNold webtool. Our results demonstrated that RhoTermPredict is a very performing algorithm both for limited-length sequences (F 1 -score obtained about 0.7) and for a genome-wide analysis. Furthermore the degree of overlap with ARNold predictions was very low. Conclusions: Our analysis shows that RhoTermPredict is a powerful tool for Rho-dependent terminators search in the three analyzed genomes and could fill this gap in computational genomics. We conclude that RhoTermPredict could be used in combination with an intrinsic terminators predictor in order to predict all the transcription terminators in bacterial genomes.