A Computational Approach to Search for Non-Coding RNAs in Large Genomic Data

Abstract

Over the last few years several specialized software tools have been developed, each allowing a certain class of RNAs insequencedatatobe found.Herewedescribeageneral tool that allows us to specify many different non-coding RNAs and structural RNA elements by a simple pattern description language.To take into account that RNA is normally conserved in structure as well as in sequence, the pattern description language combines methods to describe sequence and structural similarities as well as further characteristics, e.g., thermodynamic constraints. Structure- and sequence-based patterns describing certain classes of RNAs are collected in a web-based pattern library. These include simple patterns, e.g., describing extrastable tetraloops and small regulatory stem-loop structures, as well as more complex patterns, for example describing pseudoknots, ribozymes, SRP RNAs, 5S RNA and selenocysteine insertion sequences.Aweb-based service allows a user to search the patterns fromthe library in sequences given by the user. Alternatively, the user can specify a pattern that is searched for in public genomic sequence data. Here we give a comprehensive introduction of the pattern language, describe how to systematically derive pattern descriptions, and show some results on purine riboswitches obtained using this computational approach.