Prediction of transcription terminators in bacterial genomes

This study describes an algorithm that finds rho-independent transcription terminators in bacterial genomes and evaluates the accuracy of its predicti ons. The algorithm identifies terminators by searching for a common mRNA mo tif: a hairpin structure followed by a short uracil-rich region. For each t erminator, an energy-scoring function that reflects hairpin stability, and a tail-scoring function based on the number of U nucleotides and their prox imity to the stem, are computed. A confidence value can be assigned to each terminator by analyzing candidate terminators found both within and betwee n genes, and taking into account the energy and tail scores. The confidence is an empirical estimate of the probability that the sequence is a true te rminator. The algorithm was used to conduct a comprehensive analysis of 12 bacterial genomes to identify likely candidates for rho-independent transcr iption terminators. Four of these genomes (Deinococcus radiodurans, Escheri chia coli, Haemophilus influenzae and Vibrio cholerae) were found to have l arge numbers of rho-independent terminators. Among the other genomes, most appear to have no transcription terminators of this type, with the exceptio n of Thermotoga maritima. A set of 131 experimentally determined E. coli te rminators was used to evaluate the sensitivity of the method, which ranges from 89% to 98%, with corresponding false positive rates of 2 % and 18 %. ( C) 2000 Academic Press.