A Multidisciplinary Approach to RNA Localisation - Biophysical Approaches to Translational Control of Gene Expression - page 216

11.2.4 RNA Secondary Structure Prediction

Traditionally, RNA secondary structure prediction algorithms, such as Mfold and
RNAfold, have only predicted canonical (plus G:U) base pairings (Zuker 1994 ;
Hofacker 2003 ). However, it is becoming clear that RNA non-canonical base pair-
ings such as those in the kink-turn motif can be crucial to their function. There are
several computational approaches capable of predicting non-canonical interactions.
MC-Fold uses “nucleotide cyclic motifs”, which are small structural fragments
from experimentally derived RNA structures, used to predict lowest free energy
secondary structures (Parisien and Major 2008 ) . The CONTRAfold approach is
also able to predict non-canonical base pairings through the implementation of
probabilistic models (Do et al. 2006 ). Conditional log-linear models (CLLMs) are a
generalised form of SCFGs, and include parameters from MFE models. This
enhancement to pure probabilistic models enables CONTRAfold to outperform
pure SCFG methods. A more recent method, RNAwolf, also predicts extended RNA
secondary structures including non-canonical base pairings (zu Siederdissen et al.
2011 ). The method is based on a statistical analysis of the frequencies of the 12
basic types of base pairings of Leontis and Westhof in experimentally derived struc-
tures (Leontis and Westhof 2001, 2002 ) . A classi fi cation of the pseudo-pairs between
nucleotide bases and amino acids in 446 nucleotide-protein complexes (including
DNA as well as RNA) was undertaken to quantify preferences in the interactions
(Kondo and Westhof 2011 ). It was found that the majority of bases interact through
canonical interactions to guanine and asparagine residues. The Hoogsteen edge was
mainly presented by adenine and guanine bases and interactions with the sugar edge
were rarely seen (Fig. 11.3a ).

11.3 Identi fi cation and Characterisation of RNA Binding

Trans -Acting Factors

Identification of RNA binding trans -acting factors can be achieved in three different
ways: First, biochemical pull down approaches using a known localising RNA as
bait typically attached to beads in a column, to fish out the proteins binding the RNA
from cell extracts. Proteins directly binding the RNA, as well as those indirectly
binding, are eluted from the column. The proteins eluted can then be identified
using mass spectrometry. One example of such a technique is GRNA chromatogra-
phy (Czaplinski et al. 2005 ). Second, carrying out a genetic screen relying on an
RNA (mis)localisation phenotype. In the case of the developing Drosophila oocyte
for example, many of the trans -acting factors implicated in the localisation of the
key axis determining mRNAs gurken , oskar , bicoid and nanos were found to belong
to a group of genes earlier identified as “maternal-effect” female sterile mutations
involved in the same process (Schupbach and Wieschaus 1989 ; Luschnig et al.
2004 ). A similar screen was also performed in zebrafish for mutations affecting