Project Abstract
RNA structure and function are intimately linked. To sort out what the myriad RNAs in transcriptomes are
doing, we need rigorous approaches that also infer structure. Many predictive algorithms already exist, but they
output models for any and every sequence, and different approaches often output different models for the
same sequence. This results in suboptimal models that permeate the field, and in ascribing structure to RNAs
that don't in fact have any conserved structure. What the field of RNA research needs now to go forward is a
computational tool that will evaluate the likelihood that a certain RNA sequence has a biological structure, and
propose that structure with the highest accuracy. The best approach to this issue has been for a long time the
comparison of homologous sequences from diverse organisms. This same approach is actually at the basis of
the most successful protein structure prediction tools. But a hindrance in the wide adoption of such approaches
for predicting RNA structure is that sequence comparison requires knowledge and expertise in computational
and structural biology as well as access to tools that are not mainstream. This proposal is about the
development of a freely available webserver for reliably predicting RNA secondary and eventually tertiary
structures using evolutionary information. This webserver will operate behind the scenes as a suite of tools (the
outputs of which will be available for interested users), from homologous sequences retrieval to evaluation of
the resulting model. First, this tool will automatically retrieve and align homologous sequences using
existing and novel algorithms. This aim will search for relevant homologs to any single sequence entered as
input, which represents an unmet challenge for most current programs. Second, the application using
covariation analysis will address the likeliness that the input RNA sequence has a conserved structure,
so not every sequence used as input will necessarily output a structure model. Subsequent modules in
the online tool will evaluate the quality of the alignment, and possibly improve this alignment, so that a model
with a confidence score could be proposed. Regardless of the outcome, the user will have a result that will take
into account evolutionary as well as up-to-date RNA structural information, so it will not be biased by the use of
a single set of parameters, as is often the case with existing predictive methods. A more holistic and
straightforward computational tool harnessing evolutionary information will help disseminate the use of those
methods to the larger RNA biology community, for maximum impact on experimental design in RNA research.