DotKnot: pseudoknot prediction using the probability dot plot under a refined energy model

Sperschneider, Jana; Datta, Amitava

We presented DotKnot, a program that detects recursive H-type pseudoknots given an RNA sequence. Pseudoknot detection is a promising and efficient approach for determining the folding of an RNA. Using pseudoknot detection tools such as DotKnot, KnotSeeker or HPknotter, one can find likely pseudoknots in a sequence with high accuracy (37, 39) . The structure of the detected pseudoknots can subsequently be investigated using laboratory or bioinformatics techniques. The remaining non-crossing sequence can be folded using secondary structure prediction algorithms in Oðn 3 Þ time and Oðn 2 Þ space. DotKnot and other pseudoknot detection approaches are very time efficient, even allowing scanning of long regions in viral genomes. DotKnot assembles pseudoknot candidates from a set of structural building blocks. This set contains stems, bulge loops, internal loops and multiloops. In general, complex pseudoknots can be constructed. However, there is a trade-off between the generality of predictable pseudoknots and the biological relevance of the result. Therefore, we restrict DotKnot to the prediction of recursive H-type pseudoknots where one of the pseudoknot stems can contain bulges and internal loops. For these recursive H-type pseudoknots, we are confident that the pseudoknot energy parameters used give a good approximation. For more complex pseudoknot folds such as those with loop-loop interactions, we would have to employ a highly assumptive energy model, thus sacrificing predictive accuracy. In the future, DotKnot will be extended to the prediction of kissing hairpins and other biologically relevant classes of pseudoknots.