Section: New Results
RNA structures
Counting pseudoknots
In a recent work published in 2004, Condon analyzed 5 recent algorithms that predict secondary structures with pseudoknots. Relying on rewriting rules, she characterized the classes of pseudoknots that may be predicted. A collaborative work [15] , [25] between Lix and Lri provides an alternative combinatorial characterization by graphs, from which enumeration follows, and, additionnally, studies a new class. In the long term, one expects to add biological constraints to these combinatorial definitions.
RNA fold and Rfam accuracy
Canonical secondary structures of RNA are those without lonely base pairs. Secondary structure prediction algorithms such as RNAfold , etc., claim to have greater accuracy in folding structures without lonely base pairs than with isolated pairs. B. Raman and P. Clote, (Relative Accuracy of RNAfold to Rfam Consensus for Canonical Secondary Structures ), validate this claim: RNAfold improves accuracy in canonical structures prediction. This is assessed by extensive experiments using RNA sequences obtained from RNA database Rfam . The accuracy of the RNAfold algorithm is evaluated with respect to the consensus secondary structure of each and every RNA family in the Rfam database. This paper also points out that for certain families in the Rfam database the consensus secondary structure is inaccurate.
Riboswitches
Towards predicting the structure of a riboswitch, the first step is to extract from the genome sequence the complete RNA sequence, that is, both the aptamer and the expression platform of the riboswitch. To predict the structure after a target molecule binds to the aptamer of the riboswitch, it is also necessary to know the sequence and in turn the structure of the expression platform: then only we could identify the subsequences of the RNA involved in an alternate, stable riboswitch structure. The second step is to predict the secondary structure with the extracted RNA sequence such that the elements of the expected riboswitch family appears in the folded secondary structure. For example, in the aptamer portion of a TPP riboswitch there is a thi -box element, whose structure, and a significant portion of the sequence as well, is conserved in Prokaryotes and in some Eukaryotes). To achieve this, it is desirable to have a database containing the correct secondary structures of known riboswitches. The Rfam database has a collection of riboswitch sequences with the consensus structure, and the sequences corresponds to just the aptamer portion. We developed a computational pipeline for generating accurate secondary structures for all TPP riboswitch entires in the Rfam database. In thiswork, we use the software tools in pipeline to achieve the following: (a) retrieve sequences from genome banks corresponding to TPP riboswitch entries in Rfam , (b) locate the aptamer portion in the retrieved sequence, and (c) fold sequences to predict secondary structures that are accurate compared to the conserved structure in known TPP riboswitches.
