Abstract. In this paper we present a more efficient algorithm for comparison of minisatellites which has complexity O(n 3 + m 3 + mn 2 + nm 2 + mn) where n and m are the lengths of the maps and n and m are the sizes of run-length encoded maps. We show that this algorithm makes a significant improvement for the real biological data, dividing the computing time by a factor 30 on a significant set of data.