Alignment of RNA secondary structures using a full set of operations

We present a new alignment algorithm for pairwise comparison of RNA secondary structures represented as arc-annotated sequences. This allows to perform biologically relevant operations on RNA structures (as pairing or unpairing nucleotides) in a natural and realistic way. We describe three variants of the algorithm: global alignment, local alignment, and motif searching. All the variants run in O(n3) space and O(n4) time in the worst case, whereas the general edit distance is known to be NP-complete (2). Incidentally, we improve the time complexity of the tree local alignment algorithm of Hochsmann et al. (9) by a factor n. Additionally, our experiments strongly suggest an average- case complexity in (n2) for our algorithms. Finally, experiments on real RNA structures are reported.