On the Monte-Carlo expectation maximization for finding motifs in DNA sequences.

Abstract

Finding conserved locations or motifs in genomic sequences is of paramount importance. Expectation maximization (EM)-based algorithms are widely employed to solve motif finding problems. The present study proposes a novel initialization technique and model-shifting scheme for Monte-Carlo-based EM methods for motif finding. Two popular EM-based motif finding algorithms are compared to the proposed method, which offers improved motif prediction accuracy on a synthetic dataset and a true biological dataset.