Experimental study of modified voting algorithm for planted (l,d)-motif problem

Abstract

We consider the planted (l,d)-motif search problem, which consists of finding a substring of length l that occurs in each s i in a set of input sequences {s 1,s t } with at most d substitutions. In this paper, we study the effect of using Balla, Davila, and Rajasekaran strategy on voting algorithm practically. We call this technique, modified voting algorithm. We present an experimental study between original and modified voting algorithms on simulated data from (9,d) to (15,d). The comparison shows that the voting algorithm is faster than its modification in all instances except the instance (15,3). We also study the effect of increasing h, which is proposed by Balla, Davila, and Rajasekaran on the modified voting algorithm. From this study, we obtained the values of the number of sequences that make the running time of modified voting algorithm less than the voting algorithm and minimum. Finally, we analyze the experimental results and give some observations according to the relations: (1) l is fixed and d is variable. (2) l is variable and d is fixed. (3) l and d are variables. (4) (l,d) is challenging. © 2010 Springer Science+Business Media, LLC.