RNIE: genome-wide prediction of bacterial intrinsic terminators

Abstract

Bacterial Rho-independent terminators (RITs) are important genomic landmarks involved in gene regulation and terminating gene expression. In this investigation we present RNIE, a probabilistic approach for predicting RITs. The method is based upon covariance models which have been known for many years to be the most accurate computational tools for predicting homology in structural non-coding RNAs. We show that RNIE has superior performance in model species from a spectrum of bacterial phyla. Further analysis of species where a low number of RITs were predicted revealed a highly conserved structural sequence motif enriched near the genic termini of the pathogenic Actinobacteria, Mycobacterium tuberculosis. This motif, together with classical RITs, account for up to 90% of all the significantly structured regions from the termini of M. tuberculosis genic elements. The software, predictions and alignments described below are available from http://github.com/ppgardne/RNIE.

Figure 1.

(A) Rho-independent termination: the RNA polymerase traverses the DNA template strand from 3′ to 5′, synthesizing the nascent RNA molecule. (B) As the polymerase nears a termination site, a G + C-rich terminator stem sequence (highlighted in blue) is transcribed. (C) Formation of a hairpin structure causes the polymerase to pause, and together with a string of unstable rU-dA bonds causes the polymerase to release from the template.

Figure 2.

Alpha benchmark. The accuracy of RNIE compared to existing methods of terminator prediction. The left figure shows a ROC plot for four independent methods. The middle figure compares the sensitivity and PPV for the four methods. The figure on the right shows the speeds for each algorithm in kilobases per second.

Figure 3.

Beta benchmark. Ideal terminator predictors will generally produce predictions that are immediately 3′ to annotated genes on native sequence and no predictions on shuffled controls. For all the test genomes in Table 1 (excluding E. coli and B. subtilis), we computed the distance to the nearest 3′ genic element, including CDSs, ncRNAs and riboswitches. This was done for both native sequences and dinucleotide shuffled control sequences with corresponding gene annotation transferred to the controls. The figure on the left shows the distribution of distances for RNIE genome and gene modes and for the TransTermHP method. Inset is a barplot showing the total number of predictions for each method on native and shuffled genomes. The figures on the right show the percentage of genes that have a predicted RIT in the region −50 to +150 from an annotated 3′-end of a CDS or ncRNA across all the genome sequences described in Table 1. The upper panel illustrates the results for the native genomes, while the lower panel illustrates results for the permuted genomes.

Figure 4.

(A) The frequency of TRITs and RITs near the terminal regions of M. tuberculosis (EMBL accession: AE000516) genic features. (B) The distribution of structural stability derived P-values for the most significant M. tuberculosis terminal regions coloured by TRIT (red), RIT (black) or unclassified (blue). (C) The secondary structure and sequence conservation of the TRIT motif as displayed by R2R (27). (D&E) Sequence logos generated for the 5′ (D) and 3′ (E) halves of an alignment of the 147 copies of TRIT in the M. tuberculosis genome.