Many important RNA molecules contain pseudoknots, which are generally excluded by the definition of the secondary structure, mainly for computational reasons. Still, most existing algorithms for secondary structure prediction are not satisfactory in results and complexities, even when pseudoknots are not allowed. We present an algorithm, called P-DCFold, for the prediction of RNA secondary structures including all kinds of pseudoknots. It is based on the comparative approach. The helices are searched recursively, from more "likely" to less "likely", using the "Divide and Conquer" approach. This approach, which allows to limit the amount of searching, is possible when only non-interleaved helices are searched for. The pseudoknots are therefore searched in several steps, each helix of the pseudoknot being selected in a different step.

P-DCFold has been applied to tmRNA and RnaseP sequences. In less than two seconds, their respective secondary structures, including their pseudoknots, have been recovered very efficiently.