Match chaining algorithms for cDNA mapping

Abstract

We propose a new algorithm called the MCCM (Match Chaining-based cDNA Mapping) algorithm that allows mapping cDNAs to the genomes efficiently and accurately, utilizing local matches called MUMs (maximal unique matches) or MRMs (maximal rare matches) obtained with suffix trees. From the MUMs (or MRMs), our algorithm selects appropriate matches which are related to the cDNA mapping. We call the selection the match chaining problem. Several O(k log k)-time algorithms are known where k is the number of the input matches, but they do not permit overlaps of the matches. We propose a new O(k log k)-time algorithm for the problem with provision for overlaps. Previously, only an O(k2)-time algorithm existed. Furthermore, we also incorporate a restriction on the distances between matches for accurate cDNA mapping. We examine the performance of our algorithm through computational experiments using sequences of the FANTOM mouse cDNA database and the mouse genome. According to the experiments, the MCCM algorithm is not only very fast, but also very accurate: We achieved > 95% specificity and > 97% sensitivity at the same time against the mapping results of the FANTOM annotators. © Springer-Verlag Berlin Heidelberg 2003.