A novel algorithm for the search of 5S rRNA genes in DNA databases: comparison with other methods and identification of new potential 5S rRNA genes

Abstract

We report here a new algorithm for the identification of 5S rRNA genes in DNA databases. Based on an improved version of the general weight matrix method, this search procedure relies on the recognition of three informative regions within 5S rRNA genes, and on the weighted evaluation of the distance between them. As an additional step, the algorithm extends the weight matrix analysis to the full-length 5S rRNA sequence. This combined strategy, which includes a fast, but poorly selective, preliminary search procedure and an auxiliary step, that is slow but highly selective, strongly reduces the number of false positive instances, yielding a total false positive rate of 0.00076%. On the other hand, 97.5% of the 1045 known 5S rRNA genes were correctly recognized by this algorithm, and 29 previously unidentified potential 5S rRNA sequences were uncovered. A detailed analysis of these candidate sequences, including prediction of 5S rRNA secondary structure and checking for the presence of transcriptional termination signals, showed that eight of them correspond to authentic 5S rRNA genes. The performance of this specialized algorithm for the detection of 5S rRNA genes was compared with that of the general hidden Markov model search procedure. Due to their utilization of different filtering rules, the two approaches proved to be highly complementary. Their combined use will thus provide a very effective tool for the detection of dispersed 5S rRNA genes, either active or inactive, in the vertebrate genome.