Query driven sequence pattern mining

Abstract

The discovery of frequent patterns present in biological sequences has a large number of applications, ranging from classification, clustering and understanding sequence structure and function. This paper presents an algorithm that discovers frequent sequence patterns (motifs) present in a query sequence in respect to a database of sequences. The query is used to guide the mining process and thus only the patterns present in the query are reported. Two main types of patterns can be identified: flexible and rigid gap patterns. The user can choose to report all or only maximal patterns. Constraints and Substitution Sets are pushed directly into the mining process. Experimental evaluation shows the efficiency of the algorithm, the usefulness and the relevance of the extracted patterns.