Scalable multi-core implementation for motif finding problem

Abstract

The motif finding problem is a key step for understanding the gene regulation and expression, drug design, disease resistance, etc. Many sequential algorithms have been proposed in the literature to find the exact motifs. Voting algorithm is one such memory and time efficient sequential solution for motif finding. In this paper, we develop a parallel version of CVoting algorithm realized using openMP. The paper evaluates this parallel algorithm on a multi-core architecture using both simulated and real datasets. The paper compares the performance against existing multi-core implementations. Our experiments show that, the scalability of our implementation is linear for all challenging instances running on different number of processors, while the scalability of other implementations varies with respect to motif length or the number of processors. The average efficiency of our parallel implementations for all instances is more than 90%.