An Unambiguous And Complete Dynamic Programming Algorithm For Tree Alignment

Cedric Chauve (1) , Julien Courtiel (1) , Yann Ponty (1, 2, 3)

1 SFU - Department of Mathematics [Burnaby]
2 AMIB - Algorithms and Models for Integrative Biology
3 LIX - Laboratoire d'informatique de l'École polytechnique [Palaiseau]

Pairwise ordered tree alignment is an important pattern matching problem, motivated by RNA secondary structure comparison for example, that can be solved efficiently using Dynamic Programming (DP). An inquiry into the multiplicity of optimal solutions, as well as the existence of potentially interesting sub-optimal solutions, naturally motivates the question of exploring the space of all, optimal and sub-optimal, tree alignments. There are well-known DP-based techniques that allow for such an exploration, but they require completeness and unambiguity of the DP scheme, i.e. the existence of a score-preserving bijection between the search space and the set of possible derivations of the DP scheme. In this paper, we present the first unambiguous and complete dynamic programming algorithm for the alignment of a pair of ordered rooted trees. Our algorithm optimally aligns two trees of size $n_1$ and $n_2$ in $\Theta(n_1 n_2 max(n_1, n_2)^2)$ time in the worst-case scenario. Assuming uniformly-drawn random trees as input, it has average-case time and space complexities in $\Theta(n_1 n_2)$.

Domaines: Bio-informatique [q-bio.QM]

Cedric Chauve, Julien Courtiel, Yann Ponty. An Unambiguous And Complete Dynamic Programming Algorithm For Tree Alignment. 2015. ⟨hal-01154030v1⟩