On the Tightness of the Alternating-Cycle Lower Bound for Sorting by Reversals

Abstract

We give a theoretical answer to a natural question arising from a few years of computational experiments on the problem of sorting a permutation by the minimum number of reversals, which has relevant applications in computational molecular biology. The experiments carried out on the problem showed that the so-called alternating-cycle lower bound is equal to the optimal solution value in almost all cases, and this is the main reason why the state-of-the-art algorithms for the problem are quite effective in practice. Since worst-case analysis cannot give an adequate justification for this observation, we focus our attention on estimating the probability that, for a random permutation of n elements, the above lower bound is not tight. We show that this probability is low even for small n, and asymptotically Θ(1/n5), i.e., O(1/n5) and Ω(1/n5). This gives a satisfactory explanation to empirical observations and shows that the problem of sorting by reversals and its alternating-cycle relaxation are essentially the same problem, with the exception of a small fraction of “pathological” instances, justifying the use of algorithms which are heavily based on this relaxation. From our analysis we obtain convenient sufficient conditions to test if the alternating-cycle lower bound is tight for a given instance. We also consider the case of signed permutations, for which the analysis is much simpler, and show that the probability that the alternating-cycle lower bound is not tight for a random signed permutation of m elements is asymptotically Θ(1/m2).