Biomedical Engineering Reference
In-Depth Information
The impact of these features on evolutionary dynamics is discussed in detail
in (63,123): a population explores the sequence space in a diffusion-like manner
along the neutral network of a viable structure. Along the fringes of the popula-
tion, novel structures are produced by mutation at a constant rate (65). Fast dif-
fusion together with perpetual innovation makes these landscapes ideal for
evolutionary adaptation (30) and sets the stage for the evolutionary biotechnol-
ogy of RNA (123).
4.
CONSERVED RNA STRUCTURES
As we have seen, even a small number of randomly placed point mutations
very likely leads to a complete disruption of the RNA structure. Secondary
structure elements that are consistently present in a group of sequences with less
than, say, 95% average pairwise identity are therefore almost certainly the result
of stabilizing selection, not a consequence of the high degree of sequence con-
servation. If selection acts to preserve structure, then this structure must have
some function. It is of considerable practical interest therefore to efficiently
compute the consensus structure of a collection of such RNA molecules.
A promising approach to this goal is the combination of the "phylogenetic"
information that is contained in the sequence co-variations and the information
on the (local) thermodynamic stability of the molecules. Such methods for pre-
dicting RNA conserved and consensus secondary structure fall into two broad
groups: those starting from a multiple sequence alignment and algorithms that
attempt to solve the alignment problem and the folding problem simultaneously.
The main disadvantage of the latter class of methods (38,39,117,133) is their
high computational cost, which makes them unsuitable for long sequences such
as 16S or 23S RNAs. Most of the alignment based methods start from thermo-
dynamics-based folding and use the analysis of sequence covariations or mutual
information for post-processing; see, e.g. (55,70,79,84,85). The converse ap-
proach is taken in (50), where ambiguities in the phylogenetic analysis are re-
solved based on thermodynamic considerations.
It is important to clearly distinguish the consensus structure of a set of RNA
sequences from the collection of structural features that are conserved among
these sequences. Whenever there are reasons to assume that the structure of the
whole molecule is conserved, one may attempt to compute a consensus struc-
ture. On the other hand, consensus structures are unsuitable when a significant
part of the molecule has no conserved structures. RNA virus genomes, for in-
stance, contain only local structural patterns (such as the IRES in Picorna vi-
ruses or the TAR hairpin in HIV). Such features can be identified with a related
approach that is implemented in the
alidot
algorithm (59,60). This program
ranks base pairs using both the thermodynamic information contained in the
base pairing probability matrix and the information on compensatory, consistent,
