Computational discovery of folded RNA domains in genomes and in vitro selected libraries

Abstract

Structured functional RNAs are conserved on the level of secondary and tertiary structure, rather than at sequence level, and so traditional sequence-based searches often fail to identify them. Structure-based searches are increasingly used to discover known RNA motifs in sequence databases. We describe the application of the program RNABOB, which performs such searches by allowing the user to define a desired motif's sequence, paired and spacer elements and then scans a sequence file for regions capable of assuming the prescribed fold. Structure descriptors of stem-loops, internal loops, three-way junctions, kissing loops, and the hammerhead and hepatitis delta virus ribozymes are shown as examples of implementation of structure-based searches.

Fig. 1

(A) Descriptor for a hairpin motif specifying the first two base-pairs and allowing one mismatch in the first four base-pairs of the helix but requiring strict Watson–Crick pairing for the last two. (B) Descriptor for an internal loop allowing for variable s1, s2, and s3 regions, and one mismatch in h2. Note that because any nucleotide was allowed in these regions, a minimum of three nucleotides can appear at any position in the loops. The maximum is set by the sum of Ns and asterisks or the sum of Ns and the bracketed number. Here, the s1 element allows 3–11 nt of any sequence and s2 allows 3–20 nt. (C) Example of an RNABOB output for a search of the S. purpuratus genome using the internal loop descriptor with the s1 and s3 sequences shown in red. (D) Descriptor for a three-way junction with variable single-stranded regions. (E) Descriptor for a kissing loop allowing for one mismatch in the pseudoknot-forming h3 helix.

Fig. 2

(A) The discontinuous CLEC2 hammerhead ribozyme and the descriptor used to isolate it by Martick et al. [82]. The cleavage site is 3′ to the orange nucleotide and the catalytic residues are colored pink. (B) Descriptor for a type I hammerhead ribozyme subject to the same constraints but with different-length capping loops. (C) Descriptor for a type II hammerhead ribozyme that allows for elongated J1/2 and J2/3 regions and one mismatch in the P1 helix.

Fig. 3

(A) Secondary structures of the genomic HDV and the concensus mammalian CPEB3 ribozymes. (B) Structure descriptor for a minimal HDV-like self-cleaving sequence. (C) The drz-Agam-2-2 ribozyme, containing an extended J1/2 region, and (D) the RNABOB descriptor used to identify this ribozyme. Boxed nucleotides in the drz-Agam-2-2 secondary structure correspond to the explicitly defined nucleotides seen in (B).