A Novel Method for Predicting Ribosomal RNA Genes in Prokaryotic Genomes

With the increased efficiency of sequencing methods, large quantities of genomic information are quickly becoming available. To make use of this, methods are needed that can discern useful information from these vast data quantities. Such information could be the presence and position of specified genes with known properties. Due to their highly conserved sequences and their prevalence across all genomes the ribosomal RNA (rRNA) genes have a wide range of application within bioinformatics. Previous methods for predicting rRNA genes include RNAmmer and barrnap which both use an approach based on Hided Markov Models (HMM). However these methods are problematic due to a number of reason. Here we present a new method for rRNA gene prediction that uses a new, k-mer based, approach. This method provides a large improvement in specificity for predicting 5S rRNA genes as well as higher precision in pinpointing the ends of the rRNA gene. At the same time it preserves the high sensitivity of earlier methods with a decrease in running time. We also demonstrate the ability of rRNA gene predictors to find potential errors in the RefSeq annotation database. (Less)