Fast Motif Discovery Using a New Motif Extension Algorithm

In biology, proteins are modeled as a long chain of amino acids in primary structure. Generally, each protein is composed of 20 types of amino acids and the number and the arrangement of amino acids vary among different proteins. A sequence motif is a repeated pattern of consecutive amino acids in the primary structure of proteins which can provide information about some important biological features such as transcription factor binding and protein-protein interaction sites. In this paper, we proposed a new motif extension algorithm to enhance the performance of de Bruijn which is one of the recent motif discovery algorithms. The proposed algorithm receives an initial set of candidate motifs and tries to extend them to a desired length using a two-sided approach. In the proposed algorithm, the problem state is limited by a similarity threshold which is given by the user as a constraint. The algorithm for the development of candidate motifs always selects a characters whose appearance are greater than that of the specified similarity threshold. We conducted some experiments on real hardware and real inputs to evaluate our algorithm. The results showed that the proposed algorithm is at least 20 times faster than the original de Bruijn algorithm. Furthermore, the average similarity of identified motifs to the input protein family was 28% higher than the counterpart.